Temporal-based non-linear hearing aid prescription using a genetic algorithm Andrew T. Sabin, Holly Wiles, and Pamela Souza
Northwestern University, Department of Communication Sciences and Disorders
Amplitude
1
-1
-40 -60
250
500
1000 2000 Frequency (Hz)
4000
8000
-1
Time
Extract enveloppe
Mutation: 45 new genes were created by multiplying the values in a preserved gene by a random value drawn from a normal distrubtuion with mean 1 and std 0.25. Old Gene New Gene
Unmodified
dB SPL
100 50 0
Time
Gain (dB)
Compute Gain Control Signal 10
Time (sec)
Time
Compute Fitness
1 0
-1
(correlation coefficient (r) between modified and unmodified envelopes)
Gain (dB)
Re-Filter 0
-20 -40 -60
250
500
1000 2000 Frequency (Hz)
+ Output
4000
8000
Modified Envelope (dB)
2
4
.25 .5 1
2
4
8
100
.25 .5 1
2
4
.25 .5 1
2
4
8
Frequency (kHz)
0
.25 .5 1
2
4
Frequency (kHz)
8
2
4
2
4
.25 .5 1
2
4
2
4
8
Frequency (kHz)
1 .25 .5 1
2
4
.25 .5 1
2
4
40 20 2
4
60 40 20
1 .25 .5 1
2
4
8
.25 .5 1
2
4
Frequency (kHz)
Gain (dB)
40
30 20
40 20 .25 .5 1
2
4
Frequency (kHz)
10 0
.25 .5 1 2 4 8
40
40
30 20
20 10
50
50
40
40
30 20
.25 .5 1 2 4 8
30
0
.25 .5 1 2 4 8
.25 .5 1 2 4 8
30 20 10 0
.25 .5 1 2 4 8
50
50
40
40
30 20 10 .25 .5 1 2 4 8
Frequency (kHz)
.25 .5 1 2 4 8
30 20 10 0
.25 .5 1 2 4 8
Frequency (kHz)
*Input Levels of 50, 65, and 80 dB SPL
The following rule explains most of the variance in GA target gains. This rule is very similar to DSL (i/o) [5] 1. Attack and Release: Very short set to medians values: 6.0 and 10.7 ms respectively 2. Knee: Just above the normal-hearing detection threhsold set to median value = 17.4 dB SPL 3. Ratio: Proportional to the ratio of the dynamic range of the speech signal to that of an individual’s heaing. RatioF = 0.5063 x (RangeF, speech / RangeF, hearing ) + 0.8868 4. Gain: Proportional to the magnitude of the indivdual’s hearing loss GainF = 0.6032 x LossF + 14.6591
20
50
0
8
.25 .5 1 2 4 8
30
50
0
8
60
0
40
10
80
2
50
0
8
20 0
.25 .5 1 2 4 8
10 .25 .5 1
30
50
0
8
40
10
10
60
0
8
Gain (dB)
Gain (dB)
20
80
2
0
0
8
40
0
8
3
.25 .5 1
4
80
1
0
8
2
60
0
8
20
Gain (dB)
4
3
.25 .5 1
.25 .5 1
Gain (dB)
Gain (dB) 2
2
0
8
50
0
Gain (dB)
Ratio Ratio .25 .5 1
100
50
.25 .5 1
3
50
0
8
100
50
0
8
100
50
0
4
50
0
8
100
50
2
30
Gain (dB)
.25 .5 1
Knee (dB SPL)
100
0
.25 .5 1
40
10
80
2
50
Gain (dB)
8
0
8
1
100
50
4
3
50
0
8
2
20
Gain (dB)
4
.25 .5 1
40
Gain (dB)
4
0
8
60
Gain (dB)
2
Knee (dB SPL)
Release TIme (ms) 2
4
Gain (dB)
.25 .5 1
100
.25 .5 1
2
100
50
0
8
.25 .5 1
1
Gain (dB)
4
0
8
2
Gain (dB)
4
Ratio
2
50
0
2
Ratio
.25 .5 1
100
0
.25 .5 1
100
50
0
0
Knee (dB SPL)
100
Knee (dB SPL)
Release TIme (ms) Release TIme (ms)
8
Release TIme (ms)
Threshold (dB HL)
Attack TIme (ms) Attack TIme (ms) Attack TIme (ms)
4
50
50
80
60
40
20
r2 = 0.98
20 40 GA Target Gain (dB)
60
GA target gains were well predicted by a simple rule (similar to DSL (i/o) [5]) and had far more gain than NAL-NL2 [4], especially in the low frequencies.
Winner: First Generation
The GA-prescribed time constants were very fast. However, the compressor knee was very low, so the compressor behavior was primarily determined by the attack time constant. How the GA-prescribed fit influences perception and sound quality is currently unknown. Behavioral experiments evaluating these issues are underway.
Winner: Final Generation
10
fitness (r) = 0.74
2
50
3
The GA created a fit that compressed nearly the entire dynamic range of the speech signal into the dynamic range of hearing for a given individual at a given frequency.
Time
Unmodified Envelope (dB)
Amplitude
Apply Gain Control Signal
.25 .5 1
100
Gain
Target Gains GA NAL-NL2 [4]
Conclusions:
Notch-Clipping At Threshold
20 0
Peak-Reflection At UCL
0
Frequency (kHz)
Level (dB)
compressor knee
Apply Time Contstants
Threshold (dB HL)
.25 .5 1 2 4 8
Level (dB)
Modified Envelope Post Compressor
Time
Level (dB)
dB SPL
50
0
-50
4. Convergerence: Over the course of generations the best gene becomes more and more “fit.”
Threshold
-50
.25 .5 1 2 4 8
Distribution log log uniform uniform uniform
2. Fitness Measurement (all genes) The fitness measure was designed to reflect how well the shape of the temporal envelope was preserved after compression while placing the envelope between the detection threshold and the uncomfortable listeneing level (UCL).
.25 .5 1 2 4 8 0
50
0
Attack TIme (ms)
New Gene
100
0
0
100
Ratio
Ratio
Max 0.3 0.3 90 dB SPL 10 70 dB
Old Genes
Level (dB)
0
-50
100
Knee
Knee (dB SPL)
Min 0.001 0.001 10 dB SPL 1 0 dB
Level (dB)
Amplitude
1
0
Release
Release TIme (ms)
Parameter Attack Release Knee Ratio Gain
Level (dB)
-20
.25 .5 1 2 4 8
Attack TIme (ms)
1. Initialization Thie first generation had 100 genes, and each gene had 5 random values
Mating: 45 new genes were created by randomly combining values from the the preserved genes.
The Unmodified Channel-Specific
0
-50
-50
Time
6-Channel Bandpass Fitler
0
Attack
.25 .5 1 2 4 8
0
The Bandpass Filtered Signal
Within-Channel Processing
New Gene
Test Signal: 21 sec excerpt from the ISTS recording [3] played at both 50 and 80 dB SPL.
Envelope Post Compressor
Across-Channel Processing
Old Gene
UCL Gain (dB)
Across-Channel Processing
The input signal
Preservation: The 10 genes with the highest fitness were included unmodified in the next generation
Audiogram
Rule Target Gain (dB)
The Multiband Compressor
Overview: A genetic algorithm (GA) is an optimization procedure that mimics the mechanism of natural selection. Ideally, this procedure gradually converges on the optimal “gene.” Here each “gene” was an array of values corresponding to compressor settings. The GA was designed to find the “gene” that best preserved the shape of the original temporal envelope. Optimization was conducted separately for each channel.
The GA presciption for 5 hypothetical audiograms
Threshold (dB HL)
Determining the hearing aid parameter settings that optimally compensate for a patient’s hearing loss is critical for successful amplification. As hearing aid signal processing has become more complex, the number of potential parameter combinations has become intractably large. In the face of this complexity, hearing aid parameter settings are most often determined by applying prescriptive formulae to the patient’s audiogram. In nearly all cases, these prescriptive formulae are designed to optimize a value that is derived from the long-term average spectrum of speech (such as SII [e.g.,1] or loudness [e.g., 2]). While these prescriptions take into account the spectral variations in speech, they largely neglect the temporal variations. Several lines of research indicate that accurate perception of temporal variations, especially the temporal envelope, might be particularly important for speech perception in individuals with hearing loss. With this in mind, we began to explore the development of a non-linear hearing aid prescription that takes into account both the temporal and the spectral variations of speech. We used an optimization procedure known as a genetic algorithm.
3. Making the Next Generation After fitness is evaluated for all 100 genes, a new set of 100 genes is created by these 3 methods.
Simulation:
Threshold (dB HL)
The Genetic Algorithm
Threshold (dB HL)
Introduction
20
Time (sec)
30
40
5. Stopping: The algorithm was stopped when the same gene was the most fit for three consecutive generations. That gene was taken as the winner.
This work demonstrates that Genetic Algorithms can be used to determine the multi-band compressor settings that optimally preserve temporal envelope shape. However in future work, other GAs can be designed to optimize different fitness functions that incorporate more complex aspects of hearing such as the SII [6] or the non-linear growth of loundess [7]. References: [1] Byrne, D., Dillon, H., Ching, T., Katsch, R., and Keidser, G. (2001). NAL-NL1 procedure for fitting nonlinear hearing aids: characteristics and comparisons with other procedures, J Am Acad Audiol 12, 37-51. [2] Moore, B. C., Glasberg, B. R., and Stone, M. A. Development of a new method for deriving initial fittings for hearing aids with multi-channel compression: CAMEQ2-HF, Int J Audiol 49, 216-227. [3] Inga Holube, Short description of the International Speech Test Signal (ISTS) [4] The National Acoustic Laboratories NAL-NL2 prescription (unpublished) (Cornelisse et al. 1995; Moore and Glasberg 2004) [5] Cornelisse, L. E., Seewald, R. C., and Jamieson, D. G. (1995). The input/output formula: a theoretical approach to the fitting of personal amplification devices, J Acoust Soc Am 97, 1854-1864. [6] American National Standards Institute. (1993).American National Standards Methods for the Calculation of the Speech Intelligibility Index. [7] Moore, B. C., and Glasberg, B. R. (2004). A revised model of loudness perception applied to cochlear hearing loss, Hear Res 188, 70-88.
Acknowledgments: The authors thank Harvey Dillon and Scott Brewer at the National Acoustic Laboratories for the use of their NAL-NL2 fitting software. This work was supported by the Northwestern University Doctor of Audiology program and National Institute on Deafness and Other Communication Disorders grant F31DC009549 (to A. S.) and by R01 DC 0060014 (to PS)