Amazing sound performance comes out of the Blu

Each of the adaptive features listed above can have a varying effect on the amplified signal depending upon the listening situation. In a hearing instrument system designed to operate all day every day, the performance of these adaptive signal processing features needs to be adjusted for the listening situation. Hence the need for an automatic program, which can select the correct combination of features and  modify (steer) them, based on characterization of the environment, using classification (such as SoundNav) and sophisticated steering (such as Sound Conductor) to select appropriate settings for each feature. The resulting steering, based on default settings, represents average preferences of listeners as determined during clinical trials. 

Is there a better way?

When clients talk about making adjustments to sound performance there are typically three adjustment dimensions that they describe: focus, clarity of speech, and minimization noise. It should be noted that clients do not really know specifically which features to adjust, but rather, the user is describing sound performance effects that they wish to achieve. Unitron could have provided the client with controls that modify the performance of each individual adaptive feature directly, however that was not the direction chosen. Having control isn’t about controlling everything, its about controlling what really matters or makes a difference – adjustments that align with these relevant perceptual dimensions are provided. 

Figure 3

Figure 3. Scatterplot of overall signal level by estimated signal SNR. Each plot shows the distribution for all data (blue) and the distribution for the indicated classified environment (red). The data points for the selected environment are restricted to times when the class proportion was greater than 75%.

Reduced complexity in characterization of the listening scenario

One challenge with classification in this way is that there are six mixed environments (aka dimensions), which adds an element of uncertainty when making adjustments. However, these can be collapsed onto a mixed space having two dimensions: communication and complexity. The communication dimension represents the probability that a conversation is taking place. The complexity dimension is a combination of elements representing the listening situation. These elements can include: a) the number of and variety in types of sound objects, b) number and location of competing sound sources, c) steadiness or variability in background noise (do sound objects come and go or are the sound objects more persistent), and d) listening challenge/difficulty, (eg. speech to noise ratio). 

The classifier environments can be mapped onto this two dimensional, mixed space of communication/complexity (see Figure 4). For example, conversation in quiet and conversation in a small group map to low complexity/ high conversation (top left), whereas the noise (without speech) environment maps to high complexity/low communication (bottom right). Interestingly, there is a high correlation between the location of the classifier environments in the communication/complexity space and the overall level/SNR space. However, these two representations are not exactly the same, since the communication/complexity has the added benefit of being built on top of the prototypical sound types. Since the mapping to the communication/complexity space, at each point in time, includes the prototypical sound types there is a higher degree of confidence in the classification of the situation. For example, there can be a higher degree of certainty that a conversation is occurring when the sound type is speech in noise, than if just SPL and SNR was used to classify the signal. The addition of the sound types allows the mapping to go from a space that covers overall level/SNR to a space with the perceptual dimensions of communication and complexity. 

Figure 4

Figure 4. SoundNav classifier environments shown on two dimensional communication/complexity space.

An additional advantage of the communication/ complexity space is that it can be divided into “listening zones” and the appropriate signal processing in each zone becomes readily apparent. The communication/complexity space can be subdivided into four quadrants (see Figure 5) : 

1. easy conversations 
2. easy listening 
3. challenging conversations 
4. challenging listening 

Figure 5

Figure 5. Communication/complexity space subdivided into listening zones quadrants.

Variable application of signal processing within the
communication/ complexity space.

Fitting and client controls in Blu

From a technical implementation perspective, it is important that client adjustments applied in the field, and traditional fitting, in an office or clinical setting, are treated the same way. The goal of these fitting procedures is to modify the sound performance of the instrument for a given situation. Adjusting the sound performance of a hearing instrument is relatively simple when the device has a traditional single listening environment/ manual program. However, it becomes substantially more complex when the hearing instrument has a mixed environment automatic program. Figure 7 shows a conceptual map of the elements to consider. The hearing instrument automatic program will be comprised of: a) a classification system to characterize the listening scenario, b) a steering mechanism to adjust the adaptive features and c) the individual adaptive signal processing features. Our new automatic system, found in products on the Blu platform, Integra OS, brings together these aspects which contribute to the achieved sound performance. In addition, the HCP or client will want to make adjustments to the performance, and the listening scenario and the listener’s intention will both influence where and when the adjustments are to be applied in the hearing instrument. These last two questions (where and when) are challenging to resolve. 

1. Should an adjustment be applied to a specific situation (local), to all similar situations (generic local) or to all situations (global)? 
2. For how long should the adjustment be applied? Is this a momentary change that will be removed after a period of time (volatile) or should the change be applied forever (persistent). A third mechanism could be to remember when an adjustment has been applied to the same situation numerous times and to then learn to apply the adjustment automatically. 

Figure 7

Figure 7. Mental model for “fitting” automatic program.

At Unitron we have decided to take a two pronged approach to offer user adjustability. We expect performance in the automatic program to be well matched to the client’s expectations. When the wearer wishes to modify the sound performance there are two options: 

1. Stay in the automatic program and apply global volatile macro modifiers. These adjustments are applied to the automatic program, are applied in the moment and are temporary. 
2. Switch to an appropriate manual program (or optional app program) and make additional “generic local” persistent adjustments. Adjustments in the manual program are permanent, that is the next time the client selects the same manual program the previously modified settings will be applied. The assumption is that the client will select the same manual program for similar situations in which the same sound performance is desired. In this sense the manual programs are considered “generic local”, because the user can select the same program for multiple similar situations. 

Since, in everyday use, the hearing instrument is in the automatic program, the client will most likely first make an adjustment to the automatic program. In this case, we offer a simple button to provide “more”, either: 

• along the dimension of Clarity – to improve enhance speech in a conversation, or 
• along the dimension of Comfort – to reduce background noise. 

The boost buttons in the Remote Plus app for the automatic program are macro controls that adjust a variety of adaptive signal processing features and allow the user to get a quick adjustment without a need to think about what individual settings to adjust. These controls are temporary and are meant to be applied in the moment, without leaving the automatic program. The original settings are restored when the client reboots the instruments. 

If the macro controls in the automatic program do not provide satisfactory sound performance, then the client can select an optional program (or manual program) via the app that is suited for the situation. Unitron provides a range of pre-set programs to cover common situations that may pose a challenge for the listener. Each program offers a preset configuration suitable for the listening scenario and these programs can be personalized further. The user can adjust the program based on his or her individual preferences for the three dimensions of focus, enhancing speech and reducing noise, using sliders in the app. In this case, the changes are persistent and considered to be generic local, that is suitable for similar listening situations (eg. restaurant).

References

1. Danielle Glista, PhD, Robin O’Hagan, CDA, Leonard Cornelisse, MSc, Tayyab Shah, PhD, Donald Hayes, PhD, Sean Doherty, PhD, Jason Gilliland, PhD, and Susan Scollie, PhD, 2020, Combining passive and interactive techniques in tracking auditory ecology in hearing aid use, May 2020, Hearing Review online, https://www.hearingreview.com/inside-hearing/research/techniques-in-tracking-auditory-ecology-in-hearing-aid-use

2. Leonard Cornelisse, 2017, A conceptual framework to align sound performance with the listener’s needs and preferences to achieve the highest level of satisfaction with amplification, Unitron Technical Paper (2017-09 027-6249-02)