Sound classification system for hearing aids
A hearing aid includes a sound classification module to classify environmental sound sensed by a microphone. The sound classification module executes an advanced sound classification algorithm. The hearing aid then processes the sound according to the classification.
1. A hearing aid, comprising:
a microphone configured to sense an environmental sound; and
a processor including a sound classification module coupled to the microphone and configured to classify the sound by executing a classification algorithm, the processor configured to process the sound using an outcome of the classification of the sound for specified hearing assistance functions, the sound classification module configured to classify sound from the microphone using an environment classification scheme of two sequential stages having classifiers chosen from classifiers that use a Gaussian Mixture Model, feature sets each specific to a chosen classifier of the classifiers, and features comprising five to seven low level features within each of the feature sets, the classifiers, feature sets, and number of the features within the each of the feature sets selected for each stage to enhance classification performance and reduce associated computational cost, wherein at least some features are identified as features for both stages but are calculated once, such that an overall computational cost for the two stages is less than a direct summation of computational cost over the two stages.
2. The hearing aid of claim 1 , wherein the features comprise six low level features.
3. The hearing aid of claim 1 , wherein the sound classification module is configured to: classify the sound as one of music, speech, and non-speech; and
classify the sound as one of machine noise, wind noise, and other sounds in response to the sound being classified as the non-speech.
4. The hearing aid of claim 1 , wherein:
a first stage of the sound classification module is configured to classify sound from the microphone as one of music, speech, and non-speech; and
a second stage of the sound classification module is configured to classify sound from the microphone as one of machine noise, wind noise, and other sounds.
5. A hearing aid, comprising:
a microphone configured to sense an environmental sound; and
a processor including a sound classification module coupled to the microphone and configured to classify the sound by executing a classification algorithm, the processor configured to process the sound using an outcome of the classification of the sound for specified hearing assistance functions, the sound classification module configured to classify sound from the microphone using an environment classification scheme of two sequential stages having classifiers chosen from classifiers that use a Gaussian Mixture Model, feature sets each specific to a chosen classifier of the classifiers, and features within each of the feature sets, the classifiers, feature sets, and number of the features within the each of the feature sets selected for each stage to enhance classification performance and reduce associated computational cost, wherein at least some features are identified as features for both stages but are calculated once, such that an overall computational cost for the two stages is less than a direct summation of computational cost over the two stages.
6. The hearing aid of claim 5 , wherein the sound classification module is configured to classify the sound as one of music, speech, and non-speech.
7. The hearing aid of claim 6 , wherein the sound classification module is configured to further classify the sound as one of machine noise, wind noise, and other sounds in response to the sound being classified as the non-speech.
8. The hearing aid of claim 5 , wherein the feature set comprises six features.
9. The hearing aid of claim 5 , wherein the feature set comprises five to seven features.
10. The hearing aid of claim 5 , wherein the feature set comprises low level features.
11. The hearing aid of claim 5 , wherein the feature set comprises five to seven Mel-scale Frequency cepstral coefficients (MFCC).
12. A method for operating a hearing aid including a microphone and a processor, the method comprising:
sensing an environmental sound using the microphone;
classifying the sound using the processor by executing a classification algorithm including two sequential stages each having classifiers chosen from classifiers that use a Gaussian Mixture Model, feature sets each specific to a chosen classifier of the classifiers, and features within each of the feature sets, the classifiers, feature sets, and number of the features within the each of the feature sets selected for each stage to enhance classification performance and reduce associated computational cost, wherein at least some features are identified as features for both stages but are calculated once, such that an overall computational cost for the two stages is less than a direct summation of computational cost over the two stages; and
processing the sound using an outcome of the classification of the sound for specified hearing assistance functions.
13. The method of claim 12 , wherein the features comprise five to seven low level features.
14. The method of claim 12 , wherein the features comprise five to seven Mel-scale Frequency cepstral coefficients (MFCC).
15. The method of claim 12 , wherein classifying the sound comprises:
classifying the sound as one of music, speech, and non-speech; and
classify the sound as one of machine noise, wind noise, and other sounds in response to the sound being classified as the non-speech.
16. The method of claim 12 , wherein classifying the sound comprises:
classifying the sound as one of music, speech, and non-speech via a first stage; and
classifying the sound as one of machine noise, wind noise, and other sounds via a second stage.
17. The method of claim 12 , comprising selecting the advanced classification algorithm based on performance and computational cost for the classifying the sound.
18. The method of claim 17 , further comprising selecting a feature set for classifying the sound based on the performance and computational cost for the classifying the sound.
19. The method of claim 18 , further comprising selecting a number of features in the feature set based on the performance and computational cost for the classifying the sound.
20. The method of claim 19 , wherein selecting the number of features in the feature set comprises selecting five to seven features.
21. The method of claim 20 , wherein selecting the number of features in the feature set comprises selecting low level features.