Noise Reduction In Hearing Aids Research Articles

SVD-Based Optimal Filtering Technique for Noise Reduction in Hearing Aids Using Two Microphones

We introduce a new SVD-based (Singular value decomposition) strategy for noise reduction in hearing aids. This technique is evaluated for noise reduction in a behind-the-ear (BTE) hearing aid where two omnidirectional microphones are mounted in an endfire configuration. The behaviour of the SVD-based technique is compared to a two-stage adaptive beamformer for hearing aids developed by Vanden Berghe and Wouters (1998). The evaluation and comparison is done with a performance metric based on the speech intelligibility index (SII). The speech and noise signals are recorded in reverberant conditions with a signal-to-noise ratio of and the spectrum of the noise signals is similar to the spectrum of the speech signal. The SVD-based technique works without initialization nor assumptions about a look direction, unlike the two-stage adaptive beamformer. Still, for different noise scenarios, the SVD-based technique performs as well as the two-stage adaptive beamformer, for a similar filter length and adaptation time for the filter coefficients. In a diffuse noise scenario, the SVD-based technique performs better than the two-stage adaptive beamformer and hence provides a more flexible and robust solution under speaker position variations and reverberant conditions.

Adaptation algorithms for noise reduction in hearing aids based on auditory models

Noise reduction algorithms for hearing aids usually employ certain assumptions about the ambient noise and the target (speech) signal that are only met in certain acoustical situations. In order to reduce unwanted artifacts whenever the acoustical real-life situation does not meet these assumptions, two approaches are introduced and tested that classify the acoustic environment. Both approaches are motivated by auditory models and are used to control the parameters of noise-reduction schemes. The first approach uses the amplitude modulation spectrogram (AMS) in combination with a neural net to estimate the speech-to-noise ratio either on the broadband signal or within narrow bands. It can be used for single-channel (i.e., monaural) noise reduction even for fluctuating background noise and unfavorable signal-to-noise ratios. The second approach derives the ‘‘degree of interaural coherence’’ from the input signals to both ears that decreases with increasing reverberation and increasing number of active sound sources. It is combined with several binaural noise reduction techniques, i.e., directional filtering, dereverberation, and adaptive beamforming. The advantage of using these approaches will be presented and discussed on the basis of speech intelligibility and subjective quality assessment data.

Time varying frequency-domain Wiener filtering based on binaural cues: A method of hearing aid noise reduction

Environmental noises such as cocktail party babble, traffic, and restaurant clatter are a continuing source of irritation to hearing aid wearers. One approach to reducing this noise is to provide the wearer with a directional field of hearing. Sounds coming from the back and sides are attenuated, while sounds coming from the front, or look direction, are unattenuated. Several approaches to designing such systems have been attempted: directional microphones, passive microphone arrays, and classical adaptive filter microphone array beamformers. These techniques are compromised by poor performance in reverberant environments, undesirable packaging constraints, and large computational requirements. This paper describes an efficient algorithm for a two-microphone directional hearing system based on time-varying frequency-domain Wiener filtering. The frequency-dependent SNR estimates necessary to determine the time-varying frequency-dependent attenuation of the Wiener filter are based on interaural time-of-arrival and interaural magnitude differences. Two alternative formulas are presented for calculating these SNR estimates. A study is made of the frequency resolution requirements needed to achieve good results using the Wiener filtering approach. Results from clinical trials showing improvements in both objective intelligibility scores and subjective SRT scores are presented.

Optimizing the performance of the adaptive beamformer noise reduction scheme for hearing aids in everyday listening situations.

The adaptive beamformer approach for noise reduction in hearing aids is known to be very successful in anechoic chambers and other experimental situations, but the performance decreases rapidly in realistic environments. In order to optimize the adaptive beamformer for everyday use, the algorithm was analyzed by means of a novel theoretical approach, which estimates the performance as a function of several acoustic and design parameters. According to our findings, adaptive filters of at least 50 ms in length have to be used in order to achieve a gain in signal-to-noise ratio of approximately 4 to 6 dB in ordinary office-sized rooms with typical reverberation times of around 0.4 s. Furthermore, a reliable target signal detection algorithm is necessary in order to prevent the adaptation of the filter in the presence of strong target signals. An optimized version of the adaptive beamformer has been implemented in real time on a TMS320C30 digital signal processing system. Preliminary results of intelligibility tests will be presented. [Work supported by the Swiss National Research Foundation.]