Evaluation of Perceptual Load for Hearing Impaired using Wavelet Based Speech Processing

Abstract

Frequency transposition technique is a method in which energy in the high frequency region is transposed to low frequency region. Various speech enhancing schemes have tried to reduce the effect of increased spectral and temporal masking, in cases of sensorineural hearing impairment. To reduce the effect of spectral masking, schemes are proposed on the alteration of energy of certain frequency components. While synthesizing speech components using wavelet packets, as with the wavelet transforms, the energy of the synthesized components is concentrated in the identified frequency ranges. The decomposition tree based on modified wavelet packets is proposed for the frequency bands with ten frequency bands for different wavelets basis and speech intelligibility was checked. The changes in the acoustic attributes due to transformation were observed.  Fifteen English consonants were used in the VCV [vowel consonant vowel] context for recording the responses of hearing impaired. For qualitative assessment, phonetically balanced monosyllabic words in English language were used. There was a decreased response time for all the four processing schemes compared with unprocessed signal, signifying reduction in burden on perception process.