Research And Simulation Of A Novel Coding Strategy On Cochlear Implants

Cochlear implant is a kind of electronic device of recovering or getting hearing for the deaf. This device can change the sound signal into electrical signal to stimulate the auditory nerve fibers directly and excite the biological current, Which results in the formation of auditory in the cerebral cortex.For normal hearing people, the vibration of sound waves are collected by the auricle getting to the tympanic membrane through the external ear and reaching the inner ear perilymph by the auditory ossicles, which causes the vibration of the basement membrane. The cilia of Hair cells on the basilar membrane causes changes in membrane potential, thereby releasing neurotransmitters, and then generating potential change of auditory nerve fiber which Located at the bottom of the hair cells. The potential change get to the central through the spiral ganglion cells, which causes the sense of hearing. Feeling tone deaf patients have different degree of reducing and pathological changes of hair cell, but still retains a certain number of auditory nerve fibers and spiral ganglion cells, so that restoring hearing can use the electrical stimulation method.Cochlea implant is the imitation of function of the cochlear. According to the principle of cochlear place coding, a group of electrode implant into the cochlear inner through the scala tympani operation, external speech processor performs coding processing on the speech signal to generate the corresponding electrode electrical stimulation pulse sequence, and gets into the body. In vivo implanted device excites different parts of the auditory nerve which near the cochlea electrode in a weak direct current on the basis of this electrical stimulation pulse sequence, making the patient produce similar to real auditory experience, so as to achieve the purpose of restoring hearing.Now three major cochlear implants producers of the world are Cochlear company of Australia, MED-EL company of Austria and Advanced Bionics company of America. To the beginning of2010, the whole world has hundreds of thousands of deaf use cochlear implant, more than half of them are children.Speech coding is one of the key technologies of cochlear implant. Speech coding scheme which is used by artificial cochlea is broadly divided into two categories, namely Coarse feature scheme and Fine feature scheme.The envelope features program includes frequency domain envelope dominant feature extraction and temporal domain envelope dominant feature extraction. The frequency domain envelope dominant feature extracted the pitch and formant parametersis of the speech signal, then synthesize the corresponding stimulation pulses to stimulate the auditory nerve. The fundamental frequency (F0) reflects the changes in pitch, decides to stimulating pulse frequency; resonance peak is the peak of the speech signal, and generally take the first3peak (F1,F2,F3), The position of the frequency point are corresponding to the place of the stimulating electrodes, its amplitude depends on the amplitude of the stimulus pulse. The higher the frequency of stimulation, the electrode is closer to the base of cochlea, conversely, closer to the cupula cochleae. This kind of coding scheme consists of a F0/F2, F0/F1/F2and MPEAK scheme. Temporal envelope dominant feature extraction is based on the method of auditory model of the band pass filter group, the preprocessing of the speech signal through a plurality of band-pass filter, and then through a series of processing, finally obtains each electrode stimulation pulse sequence. This scheme can be divided into Analog Stimulation scheme (SAS scheme) and Pulsatile Stimulation scheme (CIS scheme, SPEAK scheme, ACE scheme). Now Cochlear implants products mainly in CIS scheme, SPEAK scheme, ACE scheme. Among them, the most prominent is the simulation of compression (CA) improvement method, which advanced by American scholar Wilson written in1991in an article on Nature, called the CIS scheme of and ACE scheme used by the Australian company Cochlear.Fine feature scheme including Spectral fine structure and Temporal fine structure. Respectively were virtual channel scheme of Advanced Bionics HiRes90K and Austria cochlear FAME scheme of Austria cochlear company.Cochlear implant speech coding method is basically just consider a mathematical transformation process of speech signal to electrical signal, and not from the point of view of human physiology. Because the vast majority of sensorineural deafness and pathological basis due to the loss of hair cells or functional defects, cochlear implant is a kind of device which to replace the hair cell function, change the sound signal into electrical signals and directly stimulate the auditory nerve fiber. The function of cochlear hair cell is converting the acoustic mechanical energy into biological electrical energy and releasing neurotransmitter to trigger action potential of auditory nerve. The hearing process is:firstly, the voice makes the cochlear hair cells produce receptor potentials(potential1)through the normal voice transmission system, and then the hair cells release neurotransmitters to make synaptic to proudce synaptic potential(potential2), finally auditory nerve fibers excited and proudce nerve impulse(potential3)and transmitted to the central auditory to produce auditory perception. From the incoming sound to the process of auditory nerve proudce nerve action potential, which has two links have biological changes (potential1and potential2), but because of the nonlinear transmission characteristics of feelings, parameter characteristics of two kinds of potential is different. Speech coding at present cochlear implants used to change sound into electricity directly stimulate the auditory nerve, the electric signals output only similar with normal auditory receptor potential(potential1), but not similar with the synaptic potential(potential2),so the auditory perception and normal hearing apparently have a certain difference. If given sound stimuli, auditory nerve responses with increasing intensity of sound, generation of action potential latency can reduce, but given electrical stimulation, as long as the suprathreshold stimuli, auditory nerve will response at the same time. Coding schemes in Cochlear implants existing ignore time coding effect synapses, as to ignore the auditory center characteristics of the time encoding, transmission and processing, and its characteristics is one of the most important characteristics of central auditory encoding information, it may be the main reason existing of that cochlear implants cannot make the patient obtain ideal hearing.The subject starting from the above issues, starting a study of cochlear implants coding, the main content is organized as follows:First, the subject of the international mainstream of the three cochlear coding schemes (CIS, SPEAK and ACE) were analyzed by MATLAB simulation, the three coding schemes can be classified as n of M filter bank scheme, after pretreatment of sound through the M filter to get the M output, the M output respectively extracted envelope to obtain the amplitude (A1, A2,... Select the N, Am),and Select the n maximum amplitude to stimulate the corresponding electrode. The m is the same with n value of the CIS scheme, take8or12. The SPEAK scheme of m value is20, n value is5-10, the average is6. The ACE scheme of m value is22, n value is1-20, the average is8. They are using pulse phase modulation, pulse stimulation of all electrodes are not overlapping, in alternate distribution, namely only one channel is stimulated in the same time, So as to avoid the mutual interference between different channel electrode.CIS stimulation rates generally choose900Hz, SPEAK scheme for electric pulse stimulation rate in the180to300Hz range, the average is250Hz,when the sound spectrum width, the more the number of the selected channel, stimulation rate will be lower, on the contrary, when the sound spectrum is narrower, the number of channels selected less, stimulation rate is higher. The channel selection method and stimulation of ACE scheme is similar with SPEAK scheme, the difference is the stimulation rate increased, choose the250,500,700,900,1200,1800and2400Hz, the average is about900Hz. The three scheme has a significant weakness, whether CIS, or SPEAK or ACE, as long as they are modulated by the number of channels and stimulation rate fixed, stimulus interval are the same. But not with real-time changes in stimulus intensity of sound, when the sound intensity becomes larger, the interstimulus interval will be shorter, conversely, stimulus intervals will become longer.Then the subject starting from the physiological aspects, combined with the stimulus of physiological mechanism of auditory physiology, taking the time code of hair cells and auditory nerve fibers form synapses into account, puts forward a kind of scheme using membrane potential integral discharge (Membrane Potential Integrate-and-Fire Program, referred to as MPIF) to improve the quality of speech coding of cochlear implants, under the premise of the original speech signal envelope extraction, take the membrane potential integral discharge to generate pulse sequence, also is the action potential as a stimulus pulse to stimulate the auditory nerve, in order to improve the sound from feeling for cochlear implant users new perspective to enhance the sound quality to obtain the auditory effect promotion. MPIF process is: the preprocessed voice has8output through8(expandable to12or16) a filter, then the8outputs were extracted to obtain the amplitude envelope (A1, A2,..., A8), then take the amplitude of nonlinear compression to get membrane potential, followed by membrane potential integral reaches the threshold will produce a stimulating, membrane potential also transiently decreased to zero for the next integral (if there are several channels reaches a threshold at the same time, then the integral time shortest channel stimulation, the membrane potential of these the channel also decreased to zero for the next integration), resulting in a series of stimulus sequence. As long as the integration time step is suitable, it can reduce the number of channels reaching the threshold at the same time, but also can avoid the mutual interference between channels electrode stimulation.Combined with physiological parameters, do the simulation in MATLAB platform for the MPIF scheme, which compared with the CIS scheme, MPIF scheme results show that the combination of the cochlea acoustic-electric conversion role and the role of synaptic time coding, As the intensity of the sound increase in the rate of stimulus generated sequence stimulus, an effective solution to the dynamic structure of the sequence of auditory information dissemination in the question of the role of the auditory center, making the stimulation pulse and time coding get together organically and effectively to ensure coherent stimulus pulse time and relevance.Finally, design the cochlear implant speech processor based on SEED_DEC138development board, this voice processor can collect sound signal through a microphone, take anti-aliasing filtering and AD converter, input OMAP-L138and make MPIF scheme processing and save the processing result. The signal acquisition and processing process is:the speech signal sample AD by the audio sampling chip PCM4201, sampling frequency is32kHz, the sampling number is24bit, data rate of single channel speech data is728kbps, and then is processed by McBSP to DSP, for transmission by EDMA method. Data storage uses ping-pong buffer architecture, when a buffer full is stored in a cache, process full cache data by the EDMA transmission complete interrupt, so with alternating treatment, we can be real-time rapid processing of the collected data. Then the sound of the collected signals for FIR band pass filter, rectifier, FIR low-pass filter to extract the envelope, stimulus sequences of8channel is treated by MPIF scheme, finally the stimulus sequence EDMA transmission are saved to the SDRAM.