Quality Evaluation of Combined Temporal and Spectral Processing for Hearing Impaired

This paper investigates the effectiveness of combined temporal and spectral enhancement methods for the hearing impaired listeners. The temporal enhancement algorithm involves the identification and enhancement of excitation source specific regions at gross and fine levels using linear prediction residual. The temporal enhancement algorithm is combined with four different spectral enhancement algorithms, namely, spectral subtraction, multi-band spectral subtraction, minimum mean square error short-time spectral amplitude (MMSE-STSA) and minimum mean square error log-spectral amplitude (MMSE-LSA). The temporal, spectral and combined temporal-spectral enhancement algorithms are applied on the speech signal, degraded under cafeteria, train, traffic and station noises for four SNR scenarios namely, -5, 0, 5 and 10 dB. The quality of enhanced speech signals is evaluated on both normal and hearing-impaired listeners. The combined temporal spectral processing algorithm showed significant improvement over individual spectral and temporal methods and for hearing impaired listeners the results are almost comparable with that of ideal binary masking approach for the cafeteria noise case.