Zerocrossing-based nonuniform sampling to deliver low-frequency fine structure cue for cochlear implant

Abstract

This paper introduces a zerocrossing-based nonuniform sampling algorithm to deliver the low-frequency (LF) fine structure (FS) cue for a cochlear implant (CI). At LF channels, the zerocrossing times of the bandpassed speech signal are used to represent the FS cue and evoke the pulses. At high frequency channels, the interleaved pulses are generated when no zerocrossing appears at LF channels. The fact that only one pulse is elicited at any moment diminishes the interaction of pulses from adjacent channels. Three experiments were conducted to assess the performance of the proposed algorithm, acoustic simulation of Mandarin phrase identification, intelligibility measure of the synthesized speech, and estimating the envelope waveform at LF channel. Compared with traditional continuous-interleaved-sampling algorithm, the proposed algorithm improves the correct rate on identifying Mandarin phrase contaminated by speech-shaped noise at low signal-to-noise ratio (SNR). At high SNR conditions, the synthesized speech by the proposed algorithm leads to an improved speech intelligibility measure. The envelope waveform at LF channel is faithfully estimated when a low lowpass filtering cutoff frequency was applied to extract the envelope waveform.