Critical Band Based Spectral Compression

Individuals with sensorineural hearing impairment often experience difficulty in understanding speech. This can be attributed to the widened auditory filters, which results in greater spectral spread across the critical bands and thus causing increased spectral masking. Off late various methods have been used to compress the signal towards the center of the critical band. In our study we have used gammatone filters to split the speech signal into various bands and each critical band was compressed along the frequency axis. 30 hearing impaired subjects participated in a perceptual judgment and speech identification task. From the results of our study we confirmed that processed speech became more intelligible. Thus our study highlights the need for devising a processing algorithm that can achieve spectral compression of signal towards the center of critical band in real-time. Such a processing algorithm when incorporated in digital hearing aids may prove to be beneficial for individuals with sensorineural hearing loss to improve their perceived speech intelligibility.