Tutorial
This tutorial shows how the mobile hearing aid prototype with the latest version (currently version 1.0) of the SD-card image can be employed with the Bluetooth remote control.
You will need the hardware with the most recent copy of the SD-card image, and you need to pair the Bluetooth controller. Most importantly, you will need a calibrated setup.
The image shows the controller configuration with version 0.9 of the SD-card image.
The default game pad layout already provides many functions.
The static feedback cancellation is measured once when the device boots. If the earplugs are moved afterwards, the feedback path might change and need to be re-measured. The RT button (Feedback) initiates a feedback path measurement with a noise signal that is played back three seconds. For a good result use this feature in a (relatively) quiet place, don't move, and don't touch the ear plugs. A bad feedback path estimation might result in audible feedback.
The Y (Record 10s) button can be used to start a binaural recording for the duration of 10 seconds. This function can be used capture soundscapes which might be used to create test signals for better hearing aid fitting.
The device can be powered off by pressing the START (Power OFF) button, which is intuitive for Windows users ;).
The output of the openMHA can be switched off using the SELECT (Mute MHA) button. This feature can be used to only hear the direct sound.
The button DPAD DOWN (Amplification OFF) enables the use of the identity plugin. In this configuration the device has as little effect as possible on sound perception, we could also call this mode "transparent" mode. If it everything would be perfect, you should not hear a difference to when not using the device.
The button DPAD UP (Amplification ON) enables the use of the multi-band compressor plugin. In this mode the amplification depends on the selection of the gaintable which can be selected/changed with the A, B, and X button.
The A (None) button loads a gain table which is filled with zeros, i.e., it provides no gain. Switching between "Amplification OFF" (identitiy plugin) and "Amplification ON" (multi-band compressor plugin) with this setting one can notice a faint difference in the spectral coloring. This is due to the additional latency of the multi-band compressor which makes the playback and direct sound interfere differently. The effect is also known as the comb filter effect.
The B (Linear) button loads a gain table which was generated with the openMHA fitting GUI using the prescription rule Linear40 for a frequency-independent hearing loss of approximately 45 dB. This amplification scheme preserves the differences in level between soft and loud sounds, i.e. soft and loud sounds are amplified equally. Hence, loud sounds are also amplified and can be perceived as too loud.
The X (Compressive) button loads a gain table which was generated with the openMHA fitting GUI using the prescription rule Plack2004 for a frequency-independent hearing loss of approximately 45 dB and afterwards reducing the gains by 10 dB. This amplification scheme strongly compresses the input dynamic, i.e., soft sounds are amplified while loud sounds are even attenuated. This makes the distinction of soft and loud sounds difficult but prevents the amplification of already loud sounds.
You can switch back and forth between the two settings in a loud environment to notice the difference. As the origin of soft sounds usually is further away from the listener as the origin of loud sounds, this amplification scheme seems to bring closer distant sound sources.
With the DPAD RIGHT (Noise ON) and DPAD LEFT (Noise OFF) buttons a threshold simulating noise can be enabled and disabled. The purpose of this function is to provide listeners with normal hearing with a reduced dynamic range similar to the remaining dynamic range of a (very specific) listener with impaired hearing. The effect is, that conversation partners don't know that you hear a noise and they will talk normally to you, which you might find difficult to understand. If you want to have a successful conversation in that situation you may ask you conversation partner to speak up or try to enable one of the amplification schemes with the buttons B and X. Hence, the noise function helps listeners with normal hearing to experience the effect of amplification in the context of reduced audibility of speech signals.
The LT (Individualization) button starts a menu which allows to individualize the device by measuring your hearing thresholds.
The gain tables which can be loaded with the B and X buttons are very generic and not really suitable for compensation of individual hearing loss.
There is no consensus on how to derive the best individual amplification scheme, which is why hearing aid users often have to return to their hearing care professional several times to let them tune their devices.
Many so called prescription rules exist which take as an input the individual hearing thresholds at different frequencies and output a gain table.
Two of these rules where named earlier: "Linear40" and "Plack2004".
Since version 0.9 of the SD-card image, the prototype can be used to measure the individual hearing thresholds and derive a gain table with a toy prescription rule.
Therefore, press the LT (Individualization) button.
The device will read out the possible options: "Individualization menu: A) Fitting, X) Reset, Y) Test sound, B) Return".
Button A will start the measurement of the individual hearing thresholds at four frequencies (500, 1000, 2000, 4000 Hz) for both ears.
After confirming the brief explanation by pressing the START button, a series of sweeps (tones that change in frequency over time) will be played and you are asked to answer of you heard a sweep or not.
Pressing the A button means "I heard a sweep" and pressing the X button means "I heard no sweep".
Once the process is finished, which might take 10 to 15 minutes depending on your speed, an initial gain table is created with the toy prescription rule explained in the following.
You may skip each measurement with the B button.
In this case, however, the gain table generation will fail.
If the hearing thresholds were successfully measured they can be read out pressing the A (Fitting) button in the individualization menu (since version 1.0 of the SD-card image).
To re-measure the hearing thresholds, reset them first pressing the X (Reset) button in the individualization menu.
The toy prescription rule has three parameters besides the individual hearing thresholds which determine the final gain table.
A classic broken stick non-linearity is used, as depicted in the left panel.
For each frequency, a reference level (R) is mapped to the hearing threshold (HT) if the hearing threshold (HT) exceeds the reference level (R).
The gain required to achieve this mapping is applied for lower input levels as well, which is why we call it the low level gain (LLG).
Because the gain is constant for lower input levels than the reference level R, the amplification is linear in this input level range.
The gain is then steadily reduced to zero for an input level which is R+LLG*F, where F is a factor greater or equal to one which determines the compression.
In the example 1.25 is used, which results in a compression ratio of 1:5.
Above this level the gain remains zero, i.e., the mapping is linear again.
With the described prescription rule the reduction of the gain always begins exactly at the hearing threshold. However, if the low level gain (LLG) is low, it might be desirable to maintain it for higher input levels and only start reducing it above the hearing threshold (HT). Also, if the low level gain (LLG) is high, it might be desirable to start reducing the gain already below the hearing threshold HT. This behavior is implemented with an additional parameter, as depicted in the right panel. The margin (M) between the hearing threshold (HT) and, e.g., 70 dB output level (an arbitrary choice assumed to be acceptable) is determined, it can be positive or negative (in this example it is positive). Then, the low level gain (LLG) is applied for input levels lower than R+M*B, where B is a factor less than one. For B=0 the gain is reduced starting at the hearing threshold, for B=1 the gain is reduced starting at 70dB, and for any value in-between, in-between. The low level gain (LLG) itself and compression is not affected by the factor B. Hence, the frequency dependent IO-functions are shifted in parallel to the diagonal towards 70dB with increasing values of B.
Based on this toy fitting rule the initial gain table is derived with the following parameter values: The reference level R, which strongly influences the low level gain (LLG), is set to 30 dB SPL at 1000 Hz and, depending on a focus parameter, lower at lower and higher frequencies (cf. reference_levels). The factor F is set to 1.35, resulting in a compression ratio of 0.35/1.35=1:3.86. The factor B is set to 0.5, resulting in a moderate shift of compressive regions towards 70 dB SPL.
Alternative values for the three main parameters which influence the gain table can be switched using the A, B, and X button when the individualization was successful, i.e. when the measurement of the individual hearing thresholds was successfully finished.
The A (Gain) button controls an offset for the reference levels (R). The default offset is 0 and the alternatives are -10 dB, resulting in higher low level gains (LLG), and +10dB, resulting in lower low level gains (LLG).
The B (Compression) button controls the factor F. The default value is 1.35, resulting in a compression ratio of 0.35/1.35=1:3.86. Alternative values are 1.125 (compression ratio 0.125/1.125=1:9) and 2.0 (compression ratio 1.0/2.0=1:2). Lower compression means higher gains because the low level gain (LLG) is not affected by the factor F.
The X (Boost) button controls the factor B. The default value is 0.5, resulting in a moderate shift of compressive regions towards 70 dB SPL. Alternative values are 1.0 and 0.00, resulting in a strong and no shift of compressive regions towards 70 dB SPL, respectively. The effect of B can be different for each frequency. Generally the low gains (above the hearing threshold) are increased (boosted) while high gains are reduced, which might result in more consistent gains across frequencies.
Hence, the fitting based on the toy prescription rule can be modified by three parameters with each three options during runtime to experiment with different settings in different acoustic environments.
You can measure your hearing thresholds with the threshold simulating noise turned on (preferably in a quiet place). The resulting gain table should then compensate the reduced audibility due to the threshold simulating noise. You may use the A, B, and X button to play with the parameters in different acoustic environments.
An incomplete list of everyday situations which could serve for testing different parameters:
- Train passing when waiting at the railway level crossing
- Ability to hear the signs for the blind at street crossings
- Ability to hear if a door opens/closes or a coin drop
- Street noise when it is raining
- Having a conversation in a quiet environment, e.g., whispering at the library
- Having a conversation in a noisy environment, e.g., in a crowded cafeteria
- Following conversations of multiple participants in a quiet environment
- Following conversations of multiple participants in an noisy environment
- Playing board games
- Talking on the phone
- Listening to music at home
- Listening to music at concerts
- Playing an instrument