Influence of a shift in frequency distribution and analysis rate on phoneme intelligibility in noisy environments for simulated bilateral cochlear implants

Abstract

A model was developed to simulate acoustically sound perception through a cochlear implant (CI), in order to evaluate the effects of a spectral shift and analysis rate on speech recognition in quiet and noise. In the current study, we considered two analysis rates, 250Hz and 500Hz, and two CI modes: Symmetric Bilateral Cochlear Implant (SBCI) and Shifted Bilateral Cochlear Implant (ShBCI). Processing and coding strategies used in this model were adapted from the Digisonic SP CI, manufactured by Neurelec.Intelligibility of speech signals processed to simulate different analysis rates and CI modes were assessed by a group of fifty normal-hearing subjects. The analysis rate was simulated by varying the overlap between successive analysis frames using a narrow band vocoder. With the SBCI mode, both ears were stimulated with the same signal (the same frequency filters were used). With the ShBCI mode, the filters were shifted in frequency (between the two ears). All the conditions were tested in quiet and in noisy environment with three different Signals to Noise Ratios (SNR). The database testing procedure used in this experimentation involved 3-phoneme words selected from the French Lafon’s lists. Speech signals were corrupted by addition of speech multi-talker babble noise.Results showed a significant effect of CI mode, of analysis rate and of SNR. The analysis rate effect was small in quiet and significant in noisy environments. The 500Hz analysis rate led to better performances than the 250Hz. Higher performances were also observed with the ShBCI mode in noisy environments. Results were mainly consistent with findings obtained from previous cochlear implant studies which suggest that CI users may perform better with a shifted bilateral stimulation in noisy environments, and with a higher analysis rate.