Spectral Splitting of Speech Using Time Varying Comb Filters to Improve Speech Perception in Sensorineural Hearing Loss Subjects

Persons with Sensorineural Hearing Loss (SNHL) experience a decrease in speech perception because of masking in a noisy environment. Spectral splitting of speech using time varying comb filters with complementary magnitude responses for binaural dichotic presentation is helpful in decreasing the effect of frequency masking of spectral components. Thus, there is an improvement in speech perception. FIR time varying comb filters with complementary magnitude responses of order 512 are designed using a frequency sampling method with 22 one-third octave bands ranging from 0 to 11 kHz. and the continuous shift in magnitude responses with time shifts selected below just a noticeable difference (JND), so that gap detection ability improves without offsetting the advantage of the spectral splitting technique. Filter performance is assessed by objective tests using PESQ (Perceptual Evaluation of Speech Quality), spectrographic analysis, and by subjective tests using Mean Opinion Score (MOS) for quality. Subjective test is investigated on normal hearing subjects by adding white noise at different SNRs. Test materials used for the evaluation are VC (Vowel Consonant) nonsense syllables and vowels. The results showed an improvement in the perception of processed speech in a noisy environment.