Cochlear Chip In The Research And Implementation Of Speech Enhancement Algorithm

Cochlear implant is the only medical electronic products that can be used to make severe depth or deaf patient to restore some hearing intuitive. The gap is large between domestic research and foreign research in this area, especially in the clinical research. The current foreign expensive cochlear chip monopoly of the domestic market.Therefore, the development of economical and practical products is domestic issues to be resolved.To improve the speech recognition rate of the cochlear implanted patients has been the direction of cochlear chips. When in a noisy environment, the speech recognition rate dropped drastically. From the view of cochlear chip having the small size, low power consumption and real time features, some low hardware implementation complexity speech enhancement algorithms are researched in this paper that are suitable for the field of cochlear chip in order to eliminate environmental noise. The research works include following aspects.1. Introduced several speech enhancement algorithms that are based the single microphone, including spectral subtraction, Wiener filtering algorithm and subspace-based speech enhancement algorithm, then analyzed their advantages and disadvantages.2. Introduced and researched several speech enhancement based the double microphones. These algorithms include the first-order differential microphone array technology, the technology of adaptive noise cancellation and the enhancement algorithm based on the phase difference. The research focused on the adaptive noise cancellation, for the problem of slow Convergence and voice leak caused by the crosstalk, a new way is given based on the SNR iteration step the algorithm, this method can effectively accelerate the convergence speed and can play the role of the speech endpoint detection to suppress impaction of the crosstalk. Taking into account the case of reverberation caused by the room reflections, the traditional speech enhancement algorithm can not effectively eliminate the impact of the reverberation, so using the advantages of the phase difference algorithm for speech enhancement, a new algorithm is proposed with combination of phase and adaptive noise cancellation in order to eliminate environmental stationary nose while suppressing a certain degree of room reverberation.3. For the combination algorithm of phase and adaptive noise cancellation and focused on the characteristics of cochlear implants chip, this paper gave the RTL synthesis design and timing simulation of the various modules of the algorithm, especially for the large consumption of resources in the algorithm, such as the256-piont FFT,256-piont IFFT and the LMS modules, a detailed structural and the RTL hardware optimization design is carried out.4. The hardware implementation of speech enhancement algorithm is validated on the Xilinx development board and the effect of the speech enhancement is analyzed by using the serial port to import data into MATLAB.