Designing VoiceUp - a Mobile Application Visualizing Vocal Activity Measured by a Wearable Device

(1)

Designing VoiceUp - a Mobile

Application Visualizing Vocal

Activity Measured by a

Wearable Device

Anna Viklund

October 19, 2015

Master’s Thesis in Interaction Technology and Design, 30 credits

Supervisor at TFE-UmU: Samuel Sonning

Examiner: Thomas Mejtoft

Ume˚

a University

Department of Applied Physics and Electronics

SE-901 87 UME˚

A

(2)

(3)

Abstract

(4)

(5)

List of Figures

1.1 VoxLog: a system developed by Sonvox that analyzes the user’s voice through

a collar with sensors. . . 2 4.1 Distribution of sex and age among survey respondents. . . 16

4.2 Survey respondents opinions regarding the statement ”I would like to gain

more knowledge and insight regarding my voice” (1 = Strongly disagree, 5 = Strongly agree). 86% of the respondents stated that they agreed by answering 4 or 5. . . 16

4.3 Survey respondents opinions regarding the statement ”It is hard to train

without support and correction” (1 = Strongly disagree, 5 = Strongly agree).

The figure shows that the opinions about training without feedback varied. . 17

4.4 Survey respondents opinions regarding the statement ”I would like concrete

evidences on how my voice develops over time” (1 = Strongly disagree, 5 = Strongly agree). 57% of the respondents stated that they agreed by answering 4 or 5. . . 18

4.5 Survey respondents opinions regarding the statement ”I worry about getting

voice problems that will affect my voice” (1 = Strongly disagree, 5 = Strongly agree). The figure shows that a distinct majority picks a side in contrast to the ones who do not have an opinion. . . 18

4.6 Screenshot from Voce Vista: a formant tuning tool for singers that includes

spectrograms, among other visual representations. Less advanced singers may find it hard to interpret and make use of these kind of representations. . . 21 4.7 A sketch that shows a sequence of social updates. . . 23

4.8 A sketch of a friend list that shows what friends are singing at the moment. . 23

4.9 A sketch that shows amount of singing, speaking and other statistics. . . 24 4.10 A sketch that shows current amount of singing, and how far away the goal

for singing is. The user can also document the day and specific events. . . 25 4.11 A sketch that shows how the user can set limits for how much he or she should

sing. The user can also decide whether to get tactile feedback or not. . . 26 4.12 A sketch that shows how the user can go back in history and get information

from previous days . . . 27

(8)

vi LIST OF FIGURES

4.13 A sketch that shows a voice profile based on voice aspects. . . 28 4.14 A sketch that shows a way of visualizing how an aspect changes its value. The

user can choose an aspect in the dropdown menu and see how that aspect changes its value as time moves forward. Comments on different days will also be shown as times moves. . . 28

4.15 A sketch that shows how an aspect value can be visualized in a graph. . . 29

4.16 The start view: a view where the user can see how much he or she has sung and spoken during the day and today’s events in a timeline. . . 31 4.17 The single event view: a view where more detailed information about a

spe-cific event is shown. . . 32 4.18 The activity view: a view that shows a sequence of events from the user and

the user’s friends. . . 33 4.19 The friend list view: a view where the users friends, and when they last used

their voices, is shown. If a friend is using his och her voice at the moment, a song or speech symbol is shown. . . 34 4.20 The friend profile view: a view where the user can see the friend’s vocal usage

the current day, and a sequence of the latest events from the friend. . . 35 4.21 The history view: a view where the user can see how the amount of singing

(9)

List of Tables

4.1 Opinions among singers. This table shows the median (1 = Strongly disagree,

5 = Strongly agree) and the IQR (interquartile range) for each statement in the questionnaire. . . 15

4.2 Primary and secondary needs among different types of singers. . . 19

4.3 The result of the function prioritization. . . 30

(10)

(11)

Chapter 1 Introduction

Humans are complicated beings, and throughout history people have strived to understand themselves better. One way to obtain self-knowledge is through self-tracking. Self-tracking is the process of collecting information about one’s behaviors, habits and thoughts, and reflect on the information to understand oneself better. Lately, sensors have become both smaller and better integrated with mobile devices [1], which has increased the use of technology in the self-tracking process. This trend of using technology to collect data about oneself has been referred to as the quantified self-movement [2], and sensing and wearable devices are getting increasingly popular.

1.1 Sonvox, VoxLog and VoiceUp

Sonvox AB1 _{in Ume˚}_{a, Sweden is an example of a company that works with wearable and}

sensing technology. Their main product is VoxLog, which is a system for long-term voice monitoring, helping users to gain insights about their voice use and improve their voice behavior. For example, VoxLog can be used for detecting situations of excessive voice load [3].

VoxLog consists of a collar connected to a device (see Figure 1.1) and a desktop appli-cation. The collar can be used with the device independently for whole days and can easily be taken on and off. The collar continuously collects voice data, including sound level of the voice and pitch, through an accelerometer and a microphone [3]. The collected data is later on transfered to a computer using the VoxLog Discovery software, where information about voice behavior is visualized in informative graphs. VoxLog Discovery also allows tagging, organizing and comparison of data [3].

VoxLog has mainly been used for research purposes, and is also aimed to speech thera-pists. However, Sonvox believes that their voice analysis technology could be relevant to a larger audience. One possible target group for Sonvox is singers. This thesis will explore a concept by Sonvox called VoiceUp. The concept is that of a mobile application that com-municates with a wearable voice measuring device and visualizes voice information in a way that helps singers.

1_{http://www.sonvox.com}

(12)

2 Chapter 1. Introduction

Figure 1.1: VoxLog: a system developed by Sonvox that analyzes the user’s voice through a collar with sensors.

1.2 Goals and Research Questions

This thesis explores whether a wearable voice measuring device could be relevant to singers, and in what ways. Sonvox wanted to know if their current technology can offer singers unique value, and in order to investigate this, a list of tasks were elaborated in consultation with the company:

– Conduct an analysis of singers’ needs

– Refine the VoiceUp concept with respect to singers’ needs

– Create a design for a mobile application that is appealing and proves the concept – Create a simple prototype that proves that the concept is technically feasible

A number of research questions were formulated, with respect to the goals. The following list describes these questions:

– What needs do singers have?

– Which types of singers have these needs?

(13)

Chapter 2 Theoretical Framework

In this chapter, terms and theories related to self-tracking are introduced. The concept of the quantified self movement is defined and the different stages of personal informatics are presented. Possible barriers one might face in the different stages of personal informatics are also considered.

2.1 The Quantified Self Movement

People have always strived to obtain self-knowledge. One way to understand oneself better is to collect information about one’s behaviors, habits, and thoughts, and reflect on the

information. This activity has been referred to as self-tracking, and is a research area

within the field of behavioral psychology which dates back to the ’70s [1]. The process of tracking one’s own behavior was traditionally a method used within behavior therapy as a part of the assessment, as well as the treatment [1]. However, human cognitive abilities are limited—something that can obstruct the self-tracking process. Reflecting by simply trying to remember information makes it difficult to see patterns and trends, and people may not have the time to constantly observe certain behaviors, such as number of steps walked [4]. This is where technology comes into play.

Today’s technology makes it possible for individuals to track almost all aspects of their daily lives, such as the number of steps walked, their heart rate and the amount of sleep. The increasing use of technology to collect data about oneself has been referred to as the quantified self movement [1]. Even if self-tracking has been a field of study for decades, the concept of using technology for tracking behavior is relatively new. Lately, self-tracking has become widely embodied in sensing applications. Sensors have become both smaller and better integrated with mobile devices [1], making it possible to use sensing applications for automating parts of the self-tracking process. Self-tracking devices and applications have become increasingly popular among the general public [2] and especially within the health-tracking domain. A survey conducted by Pew Research Center’s Internet & American Life Project showed that 69% of U.S. adults track a health indicator, such as activity or what food they eat [5].

The reasons for collecting data about oneself can differ from person to person. For exam-ple, people with chronic conditions are for obvious reasons more likely to track health-related data [5]. However, the human urge to obtain self-knowledge, and to use the knowledge to make changes in life, seems to be a recurring motivator.

(14)

4 Chapter 2. Theoretical Framework

2.2 The Different Stages of Personal Informatics

Personal informatics is a term introduced by Li et al. [4], which refers to a class of systems which helps people to collect personal information as well as to understand and explore the information. The two core aspects of personal informatics are collection and reflection. In order for a personal system to be effective, it should collect the necessary information for insightful reflection.

Li and colleagues also proposed a five-stage-based model of personal informatics com-posed of following stages [4]:

– Preparation: the stage before people are starting to collect information – Collection: when people are collecting the information

– Integration: when the collected information is prepared for reflection – Reflection: when people reflect on the information

– Action: when people decide what they are going to do with their newfound insights In order to design and develop effective personal informatics, it is important to be aware of the problems people are experiencing while using such systems [4]. Following sections describe the stages of personal informatics in more detail and consider possible barriers in the different stages.

2.2.1 Preparation Stage

The preparation stage concerns people’s motivation to collect personal information, how they decide which information to collect, and how they plan to collect it. According to Li et al. [4], barriers might occur in this stage if the tracking tool does not meet the users’ needs. Some people start using tools they stumble upon, and it might turn out that the tool does not support the information need. If the user must switch tool, problems are likely to occur. Most systems do not support exportation of data, and even if a system does, data formats often differ. The consequence can be that previously collected data must be abandoned. One solution to the problem is better planning when selecting which collection tool to use. Deciding what information to track is not always an easy task. One common pitfall, which Choe et al. mention [1], is tracking too many aspects. Tracking multiple aspects might not be a problem at the beginning of the tracking process, when the users motivation is still high, but in the long run this causes a risk for tracking fatigue. However, tracking multiple aspects for a short period of time can be useful since it might help users to distinguish relevant aspects. If the tracking tool provides early feedback and offers an automated data analysis, irrelevant aspects can be eliminated more effectively.

Another common pitfall when deciding what to track is to focus too much on tracking symptoms and outcomes, instead of the causes [1]. In order to gain valuable insights and know how to improve the outcome measures, it is important to also track triggers and

context. However, trigger and context aspects can be hard to identify. As mentioned

(15)

2.2. The Different Stages of Personal Informatics 5

2.2.2 Collection Stage

Different barriers might turn up when information is collected. Sometimes the problems are related to the tracking tool. For example, the user might not have access to the tool when an event occurs [4]. Other problems can be related to the user. Problems with lack of time and motivation are common barriers when collecting information [1]. Keeping up the motivation can be difficult since collecting information sometimes takes both time and a large amount of effort. Remembering to collect information can also be a barrier for some [4]. In addition to problems related to the tool or the user, the collection phase might involve problem related to the data, for example when data relies on subjective estimations.

Sensing automation has many advantages. For example, the mental workload is reduced and the data accuracy is increased [1]. However, there are also advantages with manual data tracking that are important to consider. When the data is tracked manually, the users are more engaged in the collection processes, resulting in increased awareness [1]. This shows that people are not only reflecting on the information after collection, but also while they collect it. In addition to this, some aspects are difficult to track automatically, the user’s mood for example. Tracking tools must therefore maintain a balance between sensing automation and manual data tracking.

2.2.3 Integration Stage

Before the user can reflect on the collected data, the data has to be prepared for reflection. This stage of personal informatics is called the integration stage, and according to Li et al. it can be seen as the distance between the collection stage and reflection stage [4]. If the user has to put in much effort to prepare the data for collection, the integration stage is long. For example, if data is collected on paper, the user might have to collect and transcribe the notes into a graphing tool in order to reflect on the information. If the integration stage is short, it means that the user has little responsibility in preparing the information. Some applications automate the integration process, and the integration stage gets short, seen from the user’s perspective. People can encounter different barriers in the integration stage. For example, transcribing data is time-consuming. Problems might also occur if data comes from multiple inputs and if the reflection takes place in multiple outputs [4].

2.2.4 Reflection Stage

For most people who track personal information, the collection and quantitation is only a small part of the tracking process. The ultimate goal, which is also the most difficult part of the process, is to reflect on the data, and thereby gain meaningful insights [1].

According to Li et al., there are two types of reflections: short-term and long-term reflection [4]. Short-term reflection is when users reflect on the information immediately after collecting it. This type of reflection makes the user aware of his or her current status, which can be valuable information. Pedometers which continuously update the total amount of steps is one example of a product which offers a short-time reflection. Long-time reflection takes place when users reflect on the information days or weeks after collecting it. This type of reflection allows the user to compare different time periods with each other, and see trends.

(16)

6 Chapter 2. Theoretical Framework

time. Other examples of problems in the reflection stage are sparse data and difficulties in interpretation.

2.2.5 Action Stage

When the user has reflected on the collected information, the next step is to decide what to do with the information. Some people have specific goals and are using tracking as a method to see progress towards the goals [4]. After the reflection phase, these people might change their behavior to match their goals. Most systems do not provide suggestions to the users on what to do with the information, which can be a barrier when it comes to applying the gained knowledge about oneself. [4].

(17)

Chapter 3 Methods

The process was planned based on the tasks and research questions presented in Chapter 1 and on interaction design practices. Sharp et al. [6] suggest that the process of interaction design essentially involves four basic activities:

1. Identifying needs and establishing requirements

2. Developing alternative designs that meet those requirement

3. Building interactive versions of the designs so that they can be communicated and assessed

4. Evaluating what is being built throughout the process

Based on these steps, a process for this thesis was formulated to fit the scope for this work:

1. Pre-study

(a) Needs analysis

(b) Specifying target group (c) Competitor analysis 2. Design phase

(a) Concept development (b) Function prioritization

(c) Developing the design 3. Prototyping

The methods used in the activities are based on theory from Maurya [7], Sharp et al. [6], Bryman [8], Lazar et al. [9], Moore [10], Berander et al. [11] and Nilsen [12]. In this chapter, each activity is described in more detail.

(18)

8 Chapter 3. Methods

3.1 User Needs

The first question in the list presented in Chapter 1 reads: – What needs do singers have?

In order to answer this question, an analysis of singers’ needs was conducted. The first step of the analysis was to identify needs among singers and get an overall picture of the situation. The next step was to verify the existence of needs by investigating their precence. The last step in the process was to learn more about the identified needs in order to understand them on a deeper lever. This section starts with an explanation of why user needs are important to consider, and continues with more detailed descriptions of how needs were identified, verified and understood.

3.1.1 The Importance of Users’ Needs

It is a fact that most startups fail [7]. What separates a successful startup from an unsuc-cessful one is not necessarily related to product development. Failure is more likely related to lack of need among the presumptive customers, which is why it is important to make sure that the users actually want the product before building it. In practice, this means to find a problem that is worth solving before working on the solution [7]. In order to do this, the users must be involved in the process from an early stage. However, to simply ask the users what they want is not a good approach, since the users themselves may be unaware of what they need. Henry Ford once said ”If I have asked the people what they wanted, they would have said faster horses” [7]. It is easy to mistake this quote as an argument for not listening to the users, but in this case the word ”faster” actually indicates that the users have unfulfilled needs. Identifying needs means getting to know and understand the users with the aim to build a solution that supports them in achieving their goals [6].

3.1.2 Identifying Users’ Needs

The method used to identify user needs was in-depth interviews. The initial goal was to get an overall picture of the situation, and therefore five voice teachers with various experience and different perspectives on singing were interviewed. At this early stage, the aim was to investigate following questions:

– What are the goals for singers?

– How do singers reach their goals today?

– Is there a need for them to reach the goals in a more efficient way?

The interviews were semi-structured, which means that pre-determined topics and ques-tions were specified and followed, but in a non-strict way. Semi-structured interviews are appropriate when the focus is relatively clear [8] but the opportunity to explore topics in depth is important to maintain [9]. This method was chosen for its flexibility and the aim of the interviews.

(19)

3.1. User Needs 9

The interviews were recorded with the application ACR1on a Samsung Galaxy S4 Mini

and transcribed into written notes. There are multiple methods for analyzing responses from interviews. One such method is to break responses into individual thoughts and ideas written on index cards, and to group them in lines with common ideas [9]. This method was used to draw conclusions regarding needs and problems.

3.1.3 Verifying User Needs

When the needs were identified through in-depth interviews, the next step was to collect quantitative data in order to draw statistical conclusions about the occurrence of needs. To do this, a survey was conducted. Surveys are one of the most frequently used research methods and can, if constructed properly, result in data with high validity [9]. Since surveys can be used to receive responses from a large number of people, they are suitable when the aim is to get an overview of a user population [9], which was the objective in this case.

The targeted respondents for the survey were people with an interest for singing. A questionnaire with a couple of background questions and seven statements was constructed and sent out to the target audience. The statements were all related to the identified needs, and responses were collected using five-level Likert items:

1. Strongly disagree 2. Disagree

3. Neither agree nor disagree 4. Agree

5. Strongly agree

The data from the survey were analyzed through bar charts, median values and interquar-tile ranges. To test statistical significance, the Chi-Square test was used. The Chi-Square test is a popular significance test for analyzing frequency counts [9]. When variables are measured through ranking scales that are not intervally distributed, the yield data cannot be approximated to normal distribution [9]. This is because the distance between data points may differ. For example, in the case of the Likert-scale, the distance between ”Strongly disagree” and ”Disagree” might differ from the distance between ”Disagree” and ”Neither agree nor disagree”. In this case, a non-parametric analysis method is needed, such as the Chi-Square test.

3.1.4 Understanding User Needs

One disadvantage with the survey as a research method is its limitations when it comes to getting detailed and deep responses [9]. Since surveys in general are self-administrated the opportunity to ask follow-up questions, or ask users to elaborate, is lost.

To get a deeper understanding of the needs, further interviews were conducted. This time four hobby singers were interviewed. The interviews were semi-structured, just like the interviews with the voice teachers. However, the level of structure was slightly higher. More specific questions about needs were asked, and possible solutions to problems were also discussed. The answers from these interviews were analysed in a similar way as the interviews with voice teachers: by breaking responses into individual thoughts and ideas written on index cards that later on were grouped in lines with common ideas.

(20)

3.2 Specifying Target Group

The second research question presented in Chapter 1 reads: – Which types of singers have these needs?

While investigating this question, one goal was to draw conclusions about which types of singers have the most severe needs, since these singers are most likely to benefit from voice analysis technology. To achieve this goal, the first step was to distinguish hobby singers from professionals. In order to investigate differences between hobby singers and professionals, material from the interviews and the survey was used. In the first interviews, song teachers with experience of different types of singers participated. Based on the interviews, some conclusions could be drawn regarding how hobby signers differs from professionals. In the survey, the targeted respondents included both hobby singers and professional singers, and their answers was compared to each other in order to see how their needs differed.

Another goal was to identify characteristics of early adopters. Early adopters are people who adopt new product concepts early in the product’s life cycle [10]. They find it easy to understand benefits of new technology and how to make use of these benefits. In their buying decisions they rely on their own intuition rather than current trends [10]. The general public can only be reached after the early adopters has been convinced [10], which makes the early adopters an important group to consider when introducing a new high-tech product on the market.

The characteristics of the early adopters were narrowed down based on the material from the interviews with song-teachers and singers. In the interviews, the concept of using voice analysis technology to help singers was shortly presented to the participants. Some of the participants could easily imagine how they would benefit from such a product, which made them potential early adopters. Some recurring characteristics among the early adopters could be identified, and the result was summed up in a list of attributes.

3.3 Competitor Analysis

The third research question presented in Chapter 1 reads:

– What ways of dealing with problems do singers currently have?

There already exists a variety of techniques and tools which singers use to deal with their problems, so in order to build a solution that offers singers unique value, this research question was important to consider. Besides alternatives such as voice teachers, there are several other applications and voice measuring devices for singers. In order answer the research question, a competitor analysis were conducted, meaning that mobile applications, desktop applications and other voice measuring devices were evaluated based on a list of pre-determined aspects that were considered relevant. The aim of the evaluation was to learn from the strengths and weaknesses of current competitors.

(21)

3.4. Concept Development 11

3.4 Concept Development

One of the tasks presented in Chapter 1 reads:

– Refine the VoiceUp concept with respect to singers’ needs

Accordingly, the next step in the process was to start working on a solution matching the users’ needs. However, it was not obvious how the needs should be fulfilled at this early stage of product development. Therefore, the first step was to create a base of concepts to work from. Three different ideas were elaborated as high-concepts. A high-concept is an idea described in one single sentence [7]. This compact way of describing an idea makes the concept easy to spread and communicate. A high-concept does not describe solutions in detail though. The high-concepts were later on presented for the staff at Sonvox. In consultation with the staff, one of the three concepts was chosen. When deciding which concepts to choose, following aspects were considered:

– How well does the concept match the users’ needs?

– How unique is the concept, based on the competitor analysis? – How well does the concept match the available technique?

Since a high-concept does not describe solutions in detail, the next step was to concretize the concept. In order to do this, concept sketches were created. The sketches described three different versions of the chosen concept. One of the ideas was chosen in consultation with the staff at Sonvox.

3.5 Function Prioritization

When the concept of VoiceUp was clear, all desirable functions were listed and prioritized. The prioritization was performed in consultation with the staff at Sonvox. When developing a product, time and resources are often limited, and many projects face the fact that not all desirable functionality can be implemented [13]. To make sure that the most essential functionality can be delivered, requirements must be prioritized. This is an important, but challenging, activity.

When prioritizing requirements, various aspects can be taken into account. Berander and Andrews [11] suggest some aspects that can be considered:

– Importance: How important is the requirement? This is a multifaceted aspect since there are several perspectives on importance.

– Penalty: What penalties are introduced if the requirement is not fulfilled?

– Cost: How high is the implementation cost? In the case of software development, cost is closely related to time since the main cost usually is staff hours.

– Risk: Whats risks are related to the requirement? How likely are the risks to occur? What impacts do risks have?

(22)

3.6 Developing the Design

One of the tasks presented in Chapter 1 reads:

– Create a design for a mobile application that is appealing and proves the concept Accordingly, a mockup was created in Sketch2_{, which is a vector-based graphic tool for}

designers. The first step in this process was to create an overall structure. The functions were grouped based on common features, and the navigation structure for the applications was set. When it was decided which functions each screen should provide, the next step was to concretize the functions by creating a simple mockup. At this stage, the focus was mainly on structure and interaction, rather than on graphical aspects. Based on the mockup, a low-fidelity prototype in paper was created. The prototype was tested by the staff at Sonvox, who gave feedback on structure and interaction. After the feedback, the prototype was refined and tested again several times. It is almost impossible to design a perfect user interface in one single attempt, which is why an iterative approach is appropriate when designing user interfaces [12].

The next step in the process was to refine the graphical elements of the mockup. To do this, some guidelines for the graphical design were elaborated, based on requests from the management at Sonvox. These guidelines concerned coloring, shapes, fonts and icons. In order for the design to match current trends, the online communities Dribble3_{and Pinterest}4

were used as sources of inspiration. The staff at Sonvox were included in some design decisions. For example, a vote on the application’s color scheme was held. In several cases, multiple versions of aspects and components were elaborated and discussed with the staff at Sonvox.

3.7 Prototyping

The last task in the list presented in Chapter 1 reads:

– Create a simple prototype that proves that the concept is technically feasible

In order to do this, the most essential functionality was implemented as a native applica-tion. The idea behind implementing a native application was to enable further development of the prototype and to show the communication between the mobile phone and the collar. Due to practical reasons, the iOS platform was chosen and an iPhone application was im-plemented using Xcode5_{. The application was implemented in Swift}6 _{since it is a modern}

language which many advantages. For example, Swift is easy to read and is a safer language compared to Objective C considering how pointers are handled.

2_{http://bohemiancoding.com/sketch}

3_{http://dribbble.com} 4_{http://se.pinterest.com/}

5_{http://developer.apple.com/xcode/}

(23)

Chapter 4 Results

4.1 User Needs

4.1.1 Identifying and Understanding User Needs

Needs can differ between various types of singers, both when it comes to the occurrence of needs and how the needs are expressed. During the first interviews with voice teachers, following target groups were identified:

– Professional singers – Hobby singers – Beginners – Voice teachers

How needs may differ between various types of singers is discussed later on. In the interviews with voice teachers, the following needs were identified:

– Singers want to get to know and explore their voices in a more efficient way – Singers sometimes consider solitary training difficult

– Singers want to see improvement

– Singers are worried about getting voice problems

These needs were further investigated in the interviews with singers. Following subsec-tions describe the needs in more detail, based on the knowledge gained from the interviews with the singers.

Get to Know the Voice

A recurring opinion among the interviewed singers was that the voice is a complex instrument in many respects. Since it is located inside the human body it is well hidden and can appear somewhat mysterious. The anatomy and physiological processes behind vocal usage are complicated and it can be difficult to fully understand how different aspects of the voice are related to each other.

(24)

14 Chapter 4. Results

Singing is a way to express oneself, and in order to do this freely singers strive to use the voice in an unfettered way, according to all interviewed. In order to do this it is necessary to know how to control all muscles involved and to understand what the consequences will be from using the voice in various ways. Since the voice is so complex, this can be a difficult and ongoing task. According to the interviewed singers, there is always more to learn about the voice.

Some singers also expressed that getting to know the voice is a matter of trusting it and feeling confident in knowing its current status. The voice is a living instrument which can change from one day to another depending on different circumstances, such as amount of sleep. Some singers find this hard to cope with, since they are dependent on their vocal functions. Singers have to know their limitations, which can be difficult when the current status differs.

Solitary Training

According to some of the interviewed, motivation to practice singing regularly can be a problem for singers. Especially for non-professionals, it can be difficult to make singing a habit. Some compared vocal training to going to the gym: from the beginning it can be hard finding motivation and making it a habit, especially if you are alone.

It is always difficult to make a correct judgment of oneself, and especially when it comes to singing. One problem is that vibrations in the skull makes it difficult to evaluate oneself in real-time. Sound can reach the hearing system in two ways: external and internal. The vibrations in the skull will reach the hearing system from within, which will affect how the sound is perceived. The interviewed singers mentioned different methods used for dealing with this problem. One common method is to form the hand around one ear, which will amplify the external sound. Another method is to record oneself and analyze the recording afterwards. However, these techniques are still limited. Even if the sound of the own voice is perceived correctly, the challenge to make a correct judgment and decide what to work on, remains.

Some of the interviewed singers mentioned that they value feedback from others highly. The feedback can come from a vocal teacher or from a friend with knowledge in singing, and will often lead to new insights. Even if feedback from others can be useful it has some drawbacks. Comments tend to be too kind or even dishonest, and sometimes different persons give different advice, which can lead to confusion. Another problem with feedback from others is that it can be difficult to know how to make use of the information. The feedback may not include advice on how to solve the problem, and even if it does it can be difficult to learn by following instructions by others. Some of the interviewed singers find it easier to learn by figuring out the solution on their own, even if this is time-consuming.

Improvement

(25)

4.1. User Needs 15

Voice Problems

According to one of the interviewed singers, all singers are more or less affected by voice problems. Since the voice organ is a part of the human body it is affected by the singer’s physical and mental health. Hoarseness and vocal nodules are two common problems that the singers mentioned. Some voice problem can be prevented by general voice care, while other problems are hard to do anything about. Since singers are dependent on their vocal functions, some worry about getting voice problems that will affect their singing.

4.1.2 Verifying User Needs

The questionnaire was answered by 56 persons interested in singing: 10 professional singers, 44 hobby singers and 2 beginners. The distribution by sex was somewhat uneven, as shown in Figure 4.1; 68% of the respondents were male and 32% were female. Besides distribution by sex, Figure 4.1 shows that a majority of the respondents were 25–30 years old, but that singers in all ages responded to the survey.

Statement Median IQR

I would like to gain more knowledge and insight regarding my

voice 5 1

It is hard to train without support and correction 3 1

I cannot see a vocal trainer as often as I would like to 4 2,5

I would like concrete evidences on how my voice develops

over time 4 2

I often lack motivation to do necessary song training 3 1

It is difficult to translate theory into practice in the case of

vocal technique 3 1

I worry about getting voice problems that will affect my voice 3 2

Table 4.1: Opinions among singers. This table shows the median (1 = Strongly disagree, 5 = Strongly agree) and the IQR (interquartile range) for each statement in the questionnaire. Figure 4.2 shows that a distinct majority want to get to know their voice better. 86% agreed to the statement ”I would like to gain more knowledge and insight regarding my voice”. In the case of this particular statement, Table 4.1.2 shows a median value corre-sponding to ”Strongly agree”. A Chi-Square test confirmed that the result was not a random occurrence with 99.9% significance. The tendency to agree to this statement was recurring regardless of the singing level.

(26)

Figure 4.1: Distribution of sex and age among survey respondents.

(27)

4.2. Conclusions Regarding The Target Group 17

Figure 4.3: Survey respondents opinions regarding the statement ”It is hard to train without support and correction” (1 = Strongly disagree, 5 = Strongly agree). The figure shows that the opinions about training without feedback varied.

A relative high percentage of singers in general (57%) want concrete evidence on how the voice develops over time. Figure 4.4 shows this tendency, but also that the answers were spread. This can also be seen in Table 4.1.2 (IQR = 2). A Chi-Square test confirmed that the result was not a random occurrence, with 99.9% significance.

The respondents showed strong opinions about voice problems, as Figure 4.5 illustrates. A distinct majority (86 %) picked a side in contrast to the ones that did not state an opinion. A Chi-Square test confirmed that this tendency was not a random occurrence, with 99% significance. A relatively high percentage of the professional singers (60%) agreed to the statement ”I worry about getting voice problems that will affect my voice”, but there was also a significant part (30%) that disagreed. The hobby singers answered somewhat differ-ently. 39% agreed to the statement and 45% disagreed. However, this possible difference between professionals and hobby singers could not be verified statistically.

4.1.3 Conclusions Regarding User Needs

Table 4.2 shows the conclusions regarding needs among singers, based on the result from the interviews and the survey. The primary need for both professionals and hobby singers is getting to know the voice in a more efficient way. Professionals already know their voice well, but would like to gain even more knowledge about its functions. Hobby singers have less knowledge in their voice and have a strong desire to explore it. The secondary needs are also similar, except in the case of solitary training, which leads to greater needs for hobby singers. Voice problems are present regardless of the singing level, but lead to greater needs for professionals.

4.2 Conclusions Regarding The Target Group

(28)

Figure 4.4: Survey respondents opinions regarding the statement ”I would like concrete evidences on how my voice develops over time” (1 = Strongly disagree, 5 = Strongly agree). 57% of the respondents stated that they agreed by answering 4 or 5.

(29)

4.2. Conclusions Regarding The Target Group 19

Professional singers Hobby singers

Primary needs 1. Want to get to know_{and explore their voice} in a more efficient way

1. Want to get to know and explore their voice in a more efficient way

Secondary needs

2. Want to see improvement 3. Are worried about

getting voice problems

2. Want to see improvement 3. Think that solitary

training can be hard 4. Are worried about

getting voice problems

Table 4.2: Primary and secondary needs among different types of singers.

subsections describes differences between professional singers and hobby singers in more detail, based on the knowledge gained from the interviews and the survey.

Professional Singers

Professional singers are quite a small target group. The pre-study showed that they are interested in getting to know their voices better, and could make use of a voice measuring device, but are also doing quite well on their own. If they were to use a voice measuring device, detailed measurements with high accuracy would be required. Since professional singers possess a large amount of prior knowledge about vocal usage, they would probably make use of measurements easily. Professionals also have the possibility to get a voice measuring device funded from their employer.

Hobby Singers

Hobby singers are a relative large target group and have severe needs. The pre-study showed that they really want to learn more about their voices, but tend to get stuck since the voice organ is complex. Compared to professionals, they do not require as detailed measurements. However, they may find it difficult to interpret and make use of the measurements. They also have a limited amount of money to spend on a voice measuring device.

Consequences

(30)

have in mind with this target group is that they might need more guidance when it comes to interpreting voice measurements.

Even if hobby singers are a more specified target group compared to singers in general, it is still a quite loosely defined group. To get more specific on the customer segment, the characteristics on the early adopters were narrowed down, as described in Section 3.2. The early adopters for VoiceUp are:

– Young – Exploring – Dedicated – Independent – Go-ahead – Pro-technology

4.3 Competitor Analysis

There exists several products on the market that measure and analyze the singing voice in different ways: both desktop applications, mobile applications and other voice measuring devices. 11 such products were evaluated based on pre-determined aspects. In this section, the result from the evaluation is summarized.

Most applications for desktops and phones can be used with any microphone, or with the in-built microphone of the device. Physical products that include sensors for voice measuring are not as common and tend to be more expensive. Many products on the market provide real-time feedback, and in some cases an opportunity to see how measurements change over time. However, the feedback that existing products offer is limited. Most products simply provide information about pitch, timbre and sound level. Such measurements only reflect a part of the singing voice. Another problem that existing products have is the presentation of the data. It is often difficult to know how to utilize the information. For example, knowing how to interpret and make use of the spectrograms in the formant tuning tool

Voce Vista1 _{is not obvious, as shown in Figure 4.6. Many products also have deprecated}

graphical interfaces.

Based on strengths and weaknesses of current competitors and prior knowledge about user needs, two aspects are important to consider in the case of VoiceUp. To offer users unique value VoiceUp should offer:

– Unique voice measurements

– A presentation of measurements that makes it easy for singers to utilize the information

4.4 Concept Development

4.4.1 High-Concepts

Following sections describe the three high-concepts that were elaborated. To make ideas clear, the concepts are somewhat exaggerated. For each concept, possible challenges are also described.

(31)

Figure 4.6: Screenshot from Voce Vista: a formant tuning tool for singers that includes spectrograms, among other visual representations. Less advanced singers may find it hard to interpret and make use of these kind of representations.

Concept 1: ”Minecraft for singers”

Minecraft2 _{is a popular adventure game where users can explore a dynamically generated}

world and use their imagination to interact with blocks in different ways. The game does not have a clear goal. It focuses on exploration and creativity.

The idea with the ”Minecraft for singers”-concept is letting the users explore their voices without specific goals in mind. The idea is to learn through experimentation and discover unknown aspects of the voice. The user does not necessarily have to use the voice in a correct way all the time. The important thing in this idea is to get to know the voice and to be aware of the consequences of using the voice in different ways. Challenges that come with this concept concern the development of the product. In practice, it may be difficult to develop a product that encourages exploration since ”exploration” is hard to interpret in the case of singing. It is also difficult to implement a tool that offers unlimited exploration. It is important that the product helps the user for a long period of time, which can be a problem since it may be limitations in what extent a voice measuring device can help the user to explore the voice.

Concept 2: ”The quantified voice”

This concept goes hand in hand with the quantified self movement described in Chapter 2. The idea is to measure all aspects of the voice that are possible to quantify, and to visualize the information in a way that makes it easy for users to see how aspects change over time. It is then up to the user to decide how to use the information. One possible problem with this concept is that it might be difficult for the users to know how to make use

(32)

of the information. Another risk with quantifying aspects of the voice is that it encourages users to strive toward goals. Too much focus on goal achievement may lead to unhealthy vocal usage.

Concept 3: ”The mobile voice coach”

This concept focuses on presenting vocal exercises to the user and providing feedback on how the exercises were performed. The instructions to the users should be clear. One possible challenge is that some singers find it hard to learn by taking instructions from others, as mentioned earlier. Another problem is that it is hard to provide feedback that matches the feedback singers get from vocal teachers. Furthermore, similar products already exist on the market, even if they have limitations.

Conclusions Regarding High-concepts

The concept chosen was Concept 2: ”The quantified voice”. The reason for choosing this concept was that the idea matches both the current self-tracking trend and the available technology. The collar makes it possible to measure the voice during a long period of time, which is a valuable and unique opportunity. A product based on this concept can offer something unique, which is less likely in the case of Concept 3. Furthermore, the idea is feasible, unlike Concept 1. The concept of letting users explore their voices in an unlimited way through a mobile application would be difficult to concretize.

4.4.2 Concept Sketches

Following sections describe the different versions of the chosen concept (Concept 2: ”Quan-tified voice”). All versions of the concept take advantage of the opportunity to measure the voice for a long period of time. The concept versions also share a similar goal—to increase the user’s motivation to voice training.

Concept 2.1: ”Social aspects”

This concept focuses on sharing information about voice behavior with friends. The first screen, seen in Figure 4.7, shows a sequence of social updates where the user can see what friends are singing. Events can also be liked or commented. Figure 4.8 shows what friends are singing at the moment, similar to Spotify’s3 _{friend list in the desktop application. It is}

also possible for the user to see top songs and other statistics, shown in Figure 4.9.

Concept 2.2: ”Diary”

This concept has similarities with a diary. It focuses on documentation and goal achieve-ment. The user can set goals for how much he or she should sing each day. Figure 4.10 shows the first screen, where the user can see the amount of singing and speaking. It is also possible for the user to document days or specific events. Figure 4.11 shows how the user can set goals and decide if the collar should give tactile feedback or not. The timeline, seen in Figure 4.12, allows the user to go back in history and see the amount of singing and speaking on previous days.

(33)

Figure 4.7: A sketch that shows a sequence of social updates.

(34)

(35)

(36)

Figure 4.11: A sketch that shows how the user can set limits for how much he or she should sing. The user can also decide whether to get tactile feedback or not.

Concept 2.3: ”Voice statistics”

This concept attempts to map different aspects of the voice and let the user see how aspects change over time. Figure 4.13 shows a voice profile based on the current status of the aspects. Figure 4.14 shows one way of seeing how an aspect has changed over time. This view has

similarities with SoundCloud4_{, where the user can listen to a song and see comments on}

specific parts of the song pop up while the song is playing. In the case of this concept, the user can choose an aspect in the dropdown menu and see how that aspect changes its value as time moves forward. The size of the circle will be affected by the aspect value. As time moves forward, comments on different days will also be shown, similar to SoundCloud. Figure 4.15 illustrates another way of seeing how an aspect value changes. In this case, the aspect value is visualized in a graph.

Conclusions Regarding Concept Sketches

The original idea was to pick one of the concept versions in consultation with the staff at Sonvox, and further develop the chosen concept. However, it was decided not to focus on one single version. Instead, the best features from all versions were chosen and included in the final concept. The final concept is therefore a mixture of all three versions, and is described in more detail in Section 4.5.1. The reason for not choosing a single version was that the staff at Sonvox thought all versions had some interesting features, and that a mixture of various features could be valuable to future customers.

(37)

(38)

Figure 4.13: A sketch that shows a voice profile based on voice aspects.

(39)

4.5. The Mobile Application 29

Figure 4.15: A sketch that shows how an aspect value can be visualized in a graph.

4.5 The Mobile Application

This section first describes the final concept in short. Then, all functions that are part of the concept are listed and prioritzed. Last of all, the resulting mockup showing the final design is presented and thoughts about interaction and graphical design are explained.

4.5.1 The Final Concept

VoiceUp measures how much the user sings and speaks with the aim to increase the user’s motivation to practice singing more regularly. The measurements are made through a collar around the user’s neck. The information is visualized in a mobile application, where the user can see the amount of singing and speaking for the current day, and also how the vocal usage has changed over time. The application can also be used for documenting specific days and events. In addition to this, VoiceUp is a social application which provides real-time information about when friends are using their voices. The user can also interact with friends through the application by liking and commenting on events. To sum it all up, VoiceUp provides:

– Voice statistics – Documentation – Social features

The final concept has some similarities with the application RunKeeper5_{. RunKeeper}

is an application that motivates people to run. Just like VoiceUp, the application includes statistics and social features. For example, an event is created in RunKeeper after a running session, and friends can like and comment on the event. Due to the similar ideas, the VoiceUp concept was internally called ”SongKeeper” during the concept development process, which is a name that captures the essence of VoiceUp.

(40)

4.5.2 Function Prioritization

The result of the function prioritization is summed up in Table 4.3. The table shows a list of all functions that was part of the concept, together with function type and priority value. The key features was considered to be the voice statistics functions. Without the voice statistics, the documentation and the social features lose their value. Furthermore, the voice statistics functions capture the value of the collar as a self-tracking tool, which was considered to be the most important thing. The functions are described in a more concrete way in the mockup (see Section 4.5.3).

Function Type Prio

Show how much the users’s voice has been used

recently, and in what way (song or speech) Statistics 1

Show vocal usage for a longer period of time Statistics 2

Show a daily overview where the user can see when

and how the voice was used Statistics 3

Provide an opportunity for the user to add

comments to days Documentation 4

Show voice information about the user’s friends Social 5

Provide real-time information about when and how

the user’s friends are using their voices Social 6

Show ”interesting events” from the user’s friends Social 7

Provide an opportunity for the user to comment

and like events Social 8

Table 4.3: The result of the function prioritization.

4.5.3 The Mockup and the Prototype

This section describes the mockup, its main views and the features in more detail. The

mockup consists of 4 main views described separately with screenshots and text. This

section also describes which parts of the mockup that was implemented. The Start View

(41)

10:24 10:34

2 8 ”Donec lobortis risus vel dui” TAL 13 MIN

11:58 12:11

2 8 ”Lorem ipsum dolor sit amet, consectetur adipiscing elit” SÅNG 14 MIN

12:22 12:36

Såhär har du använt rösten idag

Tidslinje

Hur har dagen varit?

MIN 37 TAL MIN 14 SÅNG

ons 29 april

ANNA VIKLUND

Min dag

Sketch 9:41 AM 100%

Start Aktivitet Historik Vänner

(42)

32 Chapter 4. Results Lägg till kommentar 0 gilla-markeringar ANNA VIKLUND Tal 10 min Lägg till en beskrivning SLUTTID 10:24 STARTTID 10:24 DATUM 2015-04-29 Händelse Sketch 9:41 AM 100% Lägg till kommentar

Samuel Sonning Quisque nec aliquam magna Karl Petersson Suspendisse et risus elementum

8 gilla-markeringar

ANNA VIKLUND Sång 13 min

Lorem ipsum dolor sit amet, consectetur adipiscing elit SLUTTID 12:36 STARTTID 12:22 DATUM 2015-04-29 Händelse Sketch 9:41 AM 100%

(43)

Aktivitet

Sketch 9:41 AM 100%

Kommentera

”Donec interdum nisl quis erat egestas [...]” Anna Viklund 2 kommentarer 8 gilla-markeringar Idag kl. 12:22 SÅNG13 MINUTER Kommentera

”Vestibulum vel mauris vulputate” Karl Petersson 1 kommentarer 2 gilla-markeringar Idag kl. 08:16 SÅNG7 MINUTER Kommentera

”Varius turpis nec, ornare nisl” Samuel Sonning

3 kommentarer 5 gilla-markeringar

Igår kl. 22:06 TAL 37 MINUTER

Figure 4.18: The activity view: a view that shows a sequence of events from the user and the user’s friends.

The Activity View

The activity view shows a sequence of events from the user and the user’s friends (see Figure 4.18). New events are added from the top. From this view the user can like friends’ events, for example. If tapping the event, the user is taken to a subview where more information about the specific event is shown (see Figure 4.17). However, if the event belongs to a friend the user can not add a description to it. In this case it is only possible to comment and like the event.

The Friend List View

(44)

ANNA VIKLUND

”Risus vel dui porta, eu ultrices diam” 2 h SAMUEL SONNING

”Consectetur adipiscing elit donec lobortis” KARL PETERSSON

”Lorem ipsum dolor sit amet”

Vänner

Sketch 9:41 AM 100%

Figure 4.19: The friend list view: a view where the users friends, and when they last used their voices, is shown. If a friend is using his och her voice at the moment, a song or speech symbol is shown.

usage the current day, and a sequence of latest events.

The History View

(45)

Kommentera

”Donec interdum nisl quis erat egestas [...]” Anna Viklund 2 kommentarer 8 gilla-markeringar Idag kl. 12:22 SÅNG13 MINUTER Kommentera

”Donec interdum nisl quis erat egestas [...]” Anna Viklund 2 kommentarer 8 gilla-markeringar Idag kl. 12:22 SÅNG13 MINUTER Senaste händelser Visa dag

Såhär mycket har Anna använt rösten idag

Aktivitet

37 MIN TAL 14 MIN SÅNG

Lorem ipsum dolor sit amet ANNA VIKLUND

Umeå, Sverige FÖLJ Proﬁl

Sketch 9:41 AM 100%

(46)

VISA DAG “Lorem ipsum dolor sit amet”

Minuter tal under dagen

37

Minuter sång under dagen

14

8 MA J 7 MA J 6 MA J 5 MA J 4 MA J MÅNAD VECKA DAG Historik Sketch 9:41 AM 100%

MÅNAD VECKA

DAG

Historik

Sketch 9:41 AM 100%

Timmar sång under månaden I genomsnitt 48 minuter

sång per dag

24

Timmar tal under månaden I genomsnitt 124 minuter

tal per dag

62

JUL JUN MA J APR MAR

(47)

The Prototype

The prototype focus on the top-priority function of the application, which is showing how much the user’s voice has been used recently, and in what way (song or speech). The prototype consist of one single view: the start view (which was described in Section 4.5.3). One important part of the prototype is the communication with the collar via Bluetooth. The collar sends messages to the application several times per second about the user’s voice activity (song, speech or silence). The voice activity is visualized in the circles for the total amount of song and speech (see Figure 4.16). When the user sings or speaks, the corresponding circle gets filled. If no voice activity is registered from the collar, the circles remain unfilled.

4.5.4 Interaction

Concerning the interaction, the main objective was to create a balance between automation and user control. As mentioned in Section 2.2.2, lack of time and motivation is two common barriers when collecting information about oneself. To overcome these barriers, VoiceUp collects information about the user’s voice automatically when he or she wears the collar. However, the opportunity for the user be part of the collection stage remains. For example, the user can add comment to days and events, and it is also possible for the user to edit or delete automatically generated events, as mentioned in Section 4.5.3. To sum it up, the idea was to create a product that not requires active interaction from the user, but provide an opportunity for the user to be part of the collection stage in order to maintain a sense of user control.

In Section 2.2.3, it was mentioned that information often has to be processed an pre-pared before the user can reflect on it and draw conclusions. When using sensors to collect information, the data quantity grows, and it can be challenging to overview and interpret the data. To make this easier, the data is heavily processed before it is presented to the user. The product classifies all voice usage as either singing or speaking and groups the information in events, which are visualized in the timeline (see Figure 4.16). As for how the amount of singing and speaking has changed over time, the goal was to present information in a way that facilitates comparison. The idea is letting the user control the level of detail by switching between day, week and month in the history view (see Figure 4.21). The amount of singing and speaking is visualized in staples, making it easier for the user to quickly get an idea of how much the voice was used.

4.5.5 Graphical Design

In consultation with the staff at Sonvox, a few guidelines for the graphical design were elaborated:

– Clean and simple appearance – Light impression

– Limited amount of colors – Thin lines and fonts

(48)

”SONG-YELLOW” (250, 208, 104)

”SPEECH-GREEN” (154, 201, 155)

”SUPER LIGHT GRAY”

(249, 249, 249) (230, 230, 230)”LIGHT GRAY”

”MIDDLE GRAY”

(164, 164, 164) ”DARK GRAY”(74, 74, 74)

Figure 4.22: The complete color scheme for the application.

interface with an appearance that competes with the content. To keep the appearance clean, the amount of elements is minimized so that only the most important ones are shown. As for adornments, they are used in a subtle way. Plenty of space around and between elements (negative space) is used in order to simplify the interface and make important content easier for the user to notice. For parts of the application which require a lot of screen area, such as the background, light colors like white and gray are used. As for other colors, it was decided to only use two colors: yellow for symbolizing song and green for symbolizing speech. This contributes to a consistent visual theme and makes it easier for the user to interpret the information. For example, the color coding makes it possible for the user to quickly distinguish a song event from a speech event (see Figure 4.16). The complete color scheme for the application is shown in Figure 4.22.

As for lines and fonts, they are consistently thin. The collection of icons that was created for the application (see Figure 4.23) is one example of how thin lines are used in the design. For the text, different versions of the font Open Sans6 _{are used due to the font’s readability}

and clean appearance. Open Sans is a sans serif typeface designed by Steve Matteson which is optimized for print, web, and mobile interfaces.

(49)

SÅNG TAL REDIGERA JUSTERA

START AKTIVITET VÄNNER HISTORIK

GILLA SKICKA INSTÄLLNINGAR NOTIS

(50)

(51)

Chapter 5 Discussion

5.1 Pre-Study

When working with a product aimed at a specific target group like singers, it is important to get to know the users and understand their needs at an early stage of the work process. The work began with studying written articles about singing. Even if the articles provided some relevant information, the obtained knowledge was not as valuable as the interviews, that were conducted later on. The conclusion is that talking to the target group should be done as early as possible, since this is the most effective way to understand their needs. Furthermore, the information gained from the interviews can often contain crucial facts which are important to be aware of at an early stage of the work.

In this pre-study, the survey was designed after the first interviews were conducted. The knowledge gained from the interviews made it easier to formulate relevant questions for the survey. Designing a survey that will contribute information is difficult without a basic understanding of the problems to be investigated. It is therefore a good idea to design the survey after a basic understanding has been obtained.

When investigating users’ needs, it may be difficult knowing which needs to focus on. Sometimes the primary needs are immediately understood and communicated by the target group. But in other situations, the target group may have several needs which are not fulfilled. In these cases it is not obvious if the primary need is the most frequent one, or the need that causes users most serious problems. In this work, needs have been prioritized mainly based on how often they occur, but in the end we decided to focus on the needs that could be met in the best way with available technology.

During the later interviews with singers, solutions to problems were discussed to some extent. However, no complete investigation of concepts hade been done at this stage. It later showed that most of the solutions brought up in the interviews were irrelevant for various reasons. For example, it showed that some solutions did not match the available technology. If we would have made decisions on concepts earlier on, we could have discussed the final concept with the singers instead. It would have been a more effective use of the time with the singers.

Designing VoiceUp - a Mobile Application Visualizing Vocal Activity Measured by a Wearable Device