Noise Reduction Schemes for Digital Hearing Aids and Their Use for the Hearing Impaired
Mark Marzinzik · Carl von Ossiezky University of Oldenburg · 2001
The aim of this thesis is to improve both the assessment methods and the available algorithms for noise reduction in hearing aids. A speech pause detection algorithm is proposed which maintains a low false-alarm rate over a wide range of signal-to-noise ratios. This facilitates its application for noise estimation in noise reduction algorithms. Comprehensive evaluations of the Ephraim-Malah noise reduction algorithms with hearing-impaired subjects show that besides better 'sound quality', most obvious benefits are reductions in the mental effort needed to listen to speech in noise. To assess this feature, a listening effort test is developed. For further subjective quality assessment of the algorithms, the method of paired comparisons in combination with the Bradley-Terry scaling model is suggested and applied. The predictive power of several 'objective' speech quality measures is investigated. Particularly the PMF and LAR objective quality measures reflect different subjective results.