Investigation of the Envelope and Phase Information for Improved Speech Perception Using an Acoustic Simulation Model for Cochlear Implants

Cochlear Implants (CI) are surgically implanted biomedical devices that can provide hearing to severely deaf people by direct stimulation of the auditory nerve, electrically. Current CI speech processors do not convey phase information, which is important for perceiving music and for hearing in noisy environments. This paper discusses a new acoustic simulation model for CI, that extracts the 'envelope' by continuous-interleaved sampling (CIS) principle and 'temporal fine structure' cue (phase information) by Hilbert algorithm. The dominant channels are identified from the filter bank outputs and subjected to temporal analysis. The amplitude estimates are derived through rectification and low-pass filtering. By Hilbert transform, the temporal fine structure is extracted through formant peaks in the signal which contributes for speech intelligibility. After band pass filtering the speech signal, carrier signals that are specific to the band output, are constructed by placing square pulses in the extracted fine structure which has greater energy level. The envelopes are subjected to amplitude-modulation using the carrier signal and the synthesized speech is generated. 15 phonemes are analyzed for its envelope and phase information and the simulation results indicated that this model produced improvement in the vowels and consonants identification significantly. The proposed model should be helpful for developing advanced speech processing strategies and improving the speech perception of CI users.