Method and implementation
The current chapter presents what methods have been utilized and a description of how the work has been implemented and why.
Concept development process
The concept development process was conducted to create external warning signals and to ensure that the result of the sound concepts was focusing on the requirements of the customers.
Identifying customer needs
To identify customer expectations on the product, interviews were conducted with customers and several potential consumers. The raw data collected from the interviews were interpreted into customer needs by following the guidelines stated in chapter 2.3.
The needs are listed below in a hierarchical order:
The warning signal is easy to hear
The warning signal indicates the behavior of the vehicle
The warning signal is harmonized with the environment
The warning signal is suitable for electric and hybrid vehicles
The warning signal is original and unique
Establishing target specifications
After identifying the customer needs, the needs were translated into metrics in order to provide a precise description of what the product has to do. Each metric was specified with a unit, a target value and rated from 1 to 5 based on its importance (see Table 2). The values in the list of metrics are based on the theoretical background and the metrics that could not be quantified were given the unit ‘subjective’.
Based on the customer needs and the target specifications, a set of product concepts was generated using the brainwriting method 6-3-5 and a concept combination table.
The focus for the 6-3 -5 method was to generate a big amount of ideas of different types of sounds using the guidelines stated in chapter 2.3. A team of five employees from the company with the titles test engineer, project manager, software engineer, task leader and system manager participated in the workshop. Each participant had to write or sketch at least three ideas every five minutes. After each five-minute round, the concepts were passed round to the adjacent participant, giving the team the opportunity to draw on each other’s ideas for inspiration. The idea generation method resulted in a total of 82 unique concepts. A selection of the ideas has been collected in a mindmap illustrated in Figure 5, and the remaining concepts can be seen in Appendix 1.
To navigate the space of possibilities, a systematic exploration was carried out using a concept combination table. The generated ideas from the internal search were organized into subcategories in order to find a wide variety of combinations (see Appendix 2). The solutions were formed by combining one fragment from different columns, creating unique sounds with diverse characteristics. Ideas that were too similar to one another were eliminated before starting the exploration process to reduce the number of arrangements.
A total of 12 combinations were created and the result of the activity is shown in the list below:
Traditional ICE + Electric sparkle + Wave + Phone vibration
Sleigh + Bees + Phone vibration + Drilling machine
Helicopter + Forest + Flute + Wave
Star ship + Bees + Ping pong
T-Ford + Bird song + Heart beat + Rain
Train + Bird song + Breeze + Flute
Forest + Classical music
Jet plane + Jungle + Downhill skiing + Coffee machine
Bike + Singing + 2000 Space odyssey + Vacuum cleaner
Boat + Screaming monkey + Drums + Saw
Cricket + Guitar + Rain + Sleigh
Laughter + Frog + Clock + Star Wars
The concept selection process was carried out using the selection method ‘product champion’. Because sounds are based on subjective sensations, the selection criteria would have been difficult to state using decision matrices for example. The evaluation of the concepts was conducted with respect to the customer needs, the measurement methods and the research question stated in chapter 1.2. The target was to find concepts that differ in rhythm and characteristic, and that could work with the design of the listening tests.
The chosen signals were:
Future: Train + Bird song + Breeze + Flute
Music: Forest + Classical music
Space: Star ship + Bee + Ping Pong + Star Wars
Electric: Traditional ICE car + Wave + Electric sparkle + Phone vibration
Nature: Cricket + Guitar + Rain + Sleigh
When the concept selection process had been finalized, sound samples were created using the music software GarageBand.
Since the sound samples are meant to alert road users, it was important to create signals that would give the proper amount of attention to pedestrians and cyclists without being annoying or causing unnecessarily dangerous levels of distraction. For example, a signal that repeats itself with small intervals are usually perceived as highly urgent but can at the same time be vastly annoying. Also, since the warning signals were going to be developed for electric and hybrid vehicles, it was important to find sounds that could fit the image of these vehicles.
To create the sound samples, synthetic instruments and effects of the software were used. The process of developing the warning signals underwent many design iterations in order to find sounds with different characteristics and rhythms. The five final signals were created with an artificial approach and are described in Table 3. All signals were low pass filtered to 2 kHz and with a frequency of 420 Hz to represent a constant speed of 10 km/h. The resolution of the sound samples was set to 16 bit.
The final step of the concept development process was to test the concepts to measure the subjective perception of the signals. The tests were carried out for four days in a music studio at Jönköping University in April, 2016.
In addition to the five sound samples presented in the previous chapter, a reference signal was included in the tests as well. The reference signal represented a Mitsubishi Colt 2010 ClearTec 1.3 and was low pass filtered at 2 kHz with an internal combustion engine at 1160 rpm, which corresponds to a speed of 10 km/h driving with 1’st gear .
A total of 48 assessors were recruited to the tests. The participants were mainly engineering students and professors aged from 19 to 60. Each test was conducted one by one over calibrated Urbanear Plattan headphones, using a Macbook Air 2012 and software from SenseLabOnline. All participants were given a short oral background description of the objective of the study, as well as written instructions of the tasks (see Appendix 3). The order of the sound samples was randomized and all signals could be looped during the presentation.
The goal of the audibility test was to find the A-weighted sound pressure level for each signal and ascertain that the warning signals are perceived equally loud as an internal combustion engine vehicle.
The simplified background noise included in the test was a pink noise signal (the power per hertz is inversely proportional to the frequency) that has been frequency weighted according to Figure 6 . The weighting curve has been constructed to fit the average spectrum of background noise measured in a number of parking lots in city and suburb environments in Denmark. The background noise was played back with an A-weighted sound pressure level of 55 dB.
A total of six assessors participated in the audibility test. The age distribution is shown in Table 4. Three different sound levels of +6 dB, +0 dB and -6 dB relative to the preliminary threshold level were tested for each warning signal, resulting in a total of 108 test rounds. Each sample sound was 10 seconds long.
The listeners were given the task to choose which sound that was different from the other two using the graphical user interface illustrated in Figure 7. When all participants had completed the test, the A-weighted sound pressure level and the dBALICE were calculated. The sound level of the reference car Lpr, was set to 50 dB(A), which represents a constant speed of 10 km/h according to the UNECE regulation.
The goal of the suitability test was to find out how external warning signals for electric and hybrid cars should sound to fit the image of the product. A total of 21 assessors participated in the test and the age distribution is shown in Table 5.
The sound samples were 15 seconds long and were presented without background noise. The listeners were given a picture of a BMW i3 as a reference electric vehicle and were asked to rate the suitability of the warning signals for the car using the graphical user interface illustrated in Figure 8. The result of the suitability test is presented in the upcoming chapter.
The goal of the annoyance test was to find out the unpleasantness of each warning signal. A total of 21 assessors participated in the test and the age distribution is shown in Table 6.
The sound samples were 15 seconds long and were presented without background noise. The listeners were asked to imagine being at home, hearing the warning signals outside as part of traffic noise and rate the annoyance of the signals using the graphical user interface illustrated in Figure 9. The result of the annoyance test is presented in the upcoming chapter.
Result and analysis
The current chapter presents the result data of this study and an analysis of the findings.
The results from the audibility test, suitability test and the annoyance test are presented below.
The obtained result of the audibility test is shown in Table 7. The second column of the table illustrates the A-weighted sound pressure levels that were assessed as the preliminary threshold levels for the warning signals in a simplified background noise. The third column shows the offset of the preliminary threshold values found as the average for the six assessors. The fourth column shows the calculated A-weighted sound pressure levels in a simplified background noise, i.e. for the stated levels all warning signals are equally audible. The fifth and last column presents the values that gives the same audibility as an internal combustion engine vehicle.
It is possible to see that for the same audibility of the warning signals; the A-weighted sound pressure levels differ more than four decibels.
Suitability and annoyance tests
The results of the suitability and annoyance tests are computed and plotted with a mean value and a 95 % confidence interval. The confidence interval may be regarded as the measuring uncertainty, which means that if the experiment is re-run there is a 95 % probability that the mean value will be within this interval. Both test results are evaluated individually but in order to find which warning signal that was considered most optimal, the differences between both test methods are used as the final test result.
The result obtained from the suitability test is illustrated in Figure 10. The warning signals are displayed on the x-axis and the numerical assessment scale (left) as well as the verbal category scale (right) are displayed on the y-axis. The confidence intervals are represented with vertical bars and the mean values with blue squares. As seen from the figure, the mean values of all sound samples are within the categories ‘Fair’ and ‘Poor’. The warning signal that was rated most suitable for electric and hybrid vehicles was the ‘Music’ signal. However, the confidence intervals show a wide distribution of opinions which means that the values of the systems may not be utterly reliable. The individual results of the suitability test are illustrated in Appendix 4.
The result obtained from the annoyance test is illustrated in Figure 11. The mean rating of all warning signals is within the verbal categories ‘Very’ and ‘Moderately’. From the figure it is possible to see that the most annoying sound sample according to the assessors was the ‘Space’ signal. The sound sample ranked as least annoying was the ‘Future’ signal, closely followed by the ‘Nature’ signal. The individual results of the annoyance test are illustrated in Appendix 5.
The confidence intervals of the annoyance test showed to be smaller for some of the sound samples compared to the result of the suitability test.
The difference between the suitability and the annoyance assessments is shown in Table 8. The result was obtained by subtracting the mean values of the annoyance test from the corresponding values of the suitability test. The signals with the highest scores are considered most optimal.
The result between the two test methods show that the distribution of the mean values is widely spread amongst the warning signals. The sound sample that received the best mean was the ‘Future’ signal and the sound sample that received the worse mean was the ‘Reference’ signal. However, the target specifications stated in chapter 3.1 indicates that the annoyance of a warning signal is weighted more important than the suitability. This means that based on the customer needs, the ‘Space’ signal is the least optimal one.
The analysis has been based on the assumption that the A- weighted sound pressure levels calculated in the audibility test are accurate. This to eliminate the possibility that the results of the suitability and annoyance tests have been affected by a difference in sound levels.
Based on the mean values, the results suggest that none of the sound samples are appropriate candidates for external warning signals. None of the samples pass the verbal category ‘Fair’ in the suitability test and none of the samples pass the verbal category ‘Moderately’ in the annoyance test. Nonetheless, it is clear from the final result displayed in Table 8 that the sound samples most preferable are those that are smooth and soft with long tone sequences. However, important to consider is that these results are solely based on listening tests. Results from field tests would perhaps appear different since surrounding objects such as buildings, road-traffic and the weather can make large impacts. As stated in chapter 2.2, sound quality is not only evaluated in terms of annoyance or suitability, it is also important with people’s interaction with the product.
Introducing new audio experiences to a group of consumers is challenging regardless of what test method is being used. Also stated in chapter 2.2 is that familiar sounds are usually perceived less annoying than unfamiliar sounds. Interesting is that the ‘Reference’ signal was rated second worse in the annoyance test although this would supposedly be the only sound that the listeners are accustomed to hearing. The large confidence interval of the signal shows however that the ratings were distributed variously amongst the listeners thus if the test were to be re-run, it might be rated less annoying. The mean values of the suitability test indicate that the ‘Reference’ signal together with the ‘Space’ signal are the least suitable warning signals for electric and hybrid vehicles such as an BMW i3
The large confidence intervals of the sound samples make it difficult to give the results reliance. Individual attributes such as mood and expectations can have a big influence in what the listeners choose to answer. The theoretical background of sound quality discusses the importance of product expectations and its relation to annoyance and unpleasantness. If the tests were to be repeated with the same assessors, the expectations of the sounds would presumably be different than before the first test round and the results would maybe change completely because of this.
Although the large confidence intervals are less favorable in order to reach a conclusion, it indicates that the educational background of the participants did not affect the outcome of the tests. However, to ensure that this supposition is true, further tests should be performed.
1.2 OBJECTIVE AND RESEARCH QUESTIONS
2 Theoretical background
2.1 QUIET ROAD TRANSPORT VEHICLES
2.2 HUMAN HEARING AND PERCEPTION OF SOUND
2.3 CONCEPT DEVELOPMENT PROCESS
2.4 MEASUREMENT METHODS
3 Method and implementation
3.1 CONCEPT DEVELOPMENT PROCESS
3.2 MEASUREMENT METHODS
4 Result and analysis
5 Discussion and conclusions
5.1 DISCUSSION OF METHOD
5.2 DISCUSSION OF FINDINGS
5.3 CONCLUSIONS AND FUTURE WORK
GET THE COMPLETE PROJECT
An initial study on external warning signals for Quiet Road Transport Vehicles