Advances in Single-Channel Noise Reduction for Hearing Instruments

Single-channel noise reduction algorithms for hearing instruments frequently require a spectral domain representation of the disturbed microphone signal. In this thesis a new block-based signal-adaptive system for short-time spectral analysis and synthesis is proposed which has two major advantages: (a) it provides a high spectral resolution and still realizes a very small system delay; (b) it allows a framewise change of the spectro-temporal properties of the spectral analysis. We show how the system can be used in a noise reduction application to improve the reproduction of stop consonants. Preference tests with normal hearing subjects and measurements of word recognition scores of subjects with cochlear implant are presented.