Self adaptive filtering of noise from speech: a hearing aid application

The authors discuss and evaluate a self-adaptive filter of noise from speech. The filter uses a single microphone, and incorporates time-series identification algorithms with artificial intelligence decision procedures to discriminate between environmental noises, including babble from speech. The filter has recently been realized in a single CMOS microchip that includes digital, analog, and ADC circuits and is presently being used by seven hearing-aid manufacturers in their otherwise conventional hearing aids, of both the behind-ear and inside-ear types. Circuit and system design considerations are discussed and performance evaluation is presented, as is the relative performance as compared with conventional nonadaptive single microphone designs based on automatic gain control and ASP-type compression circuits.<>