CN108053829A - A kind of cochlear implant coding method based on cochlea sense of hearing Nonlinear Dynamics - Google Patents

Abstract

The invention discloses an electronic cochlear encoding method based on the nonlinear dynamic mechanism of cochlear hearing, which comprises (1) establishing a nonlinear dynamic model of the cochlea; (2) constructing a nonlinear cochlear array; the nonlinear cochlear array is a group comprising n Active simulation modules with different natural frequencies, each active simulation module performs corresponding calculations on the received input voice signal, and obtains the real-time response output signal of each active simulation module; (3) outputs the real-time response of each active simulation module After the signal is rectified and low-pass filtered, the amplitude envelope is obtained; and the frequency and amplitude envelopes are converted into pulse signals and transmitted to the corresponding electrodes. The present invention utilizes the cochlear nonlinear dynamic model to enable the processed acoustic signal to exhibit a nonlinear dynamic range compression effect similar to that of the cochlear processing result, and to generate combined tones related to the pitch, thereby solving the problem of the small dynamic range of the traditional electronic cochlea. Difficult problems with tonal information transfer.

Description

A cochlear coding method based on the nonlinear dynamic mechanism of cochlear hearing

technical field

The invention belongs to the field of biomedical engineering, and more specifically relates to an electronic cochlear encoding method based on the nonlinear dynamic mechanism of cochlear hearing.

Background technique

As an important information channel for human beings to communicate with the outside world, hearing is an important way for human beings to survive and develop and perceive the world. However, a large number of people in the world live in a silent world due to various reasons. According to the statistics of the World Health Organization, there are currently 360 million people in the world who suffer from deafness or hearing impairment, accounting for 5% of the total number of people in the world. The vast majority of patients with profound deafness cannot convert the sound signal into an electrical signal that is transmitted to the auditory nerve so that they can feel the sound information. Cochlear Electronics is a human implanted hearing rehabilitation device that converts sound signals into auditory nerve signals through electrical stimulation, thereby enabling total deaf patients with complete hearing loss to regain hearing. It mainly includes two parts: a voice processor and implanted electrodes. The speech processor first converts the acoustic signal into an electrical stimulation signal on the electrodes implanted in different parts of the cochlea, and the electrical stimulation signal on these electrodes excites and induces the auditory nerve to generate a nerve electrical signal that is transmitted to the brain, so that the patient can hear the sound . At present, many brands of cochlear implants have been used clinically at home and abroad, and they have brought good news to deaf patients, allowing them to return to the world of sound. Existing cochlear implants can enable patients to obtain a certain degree of speech perception, but there is still much room for improvement in their performance. For example, cochlear implants can only provide a dynamic perception range of 20dB, while normal people can perceive sounds with a dynamic range of 120dB. Another notable problem is that cochlear implants still cannot enable patients to effectively perceive the pitch information of sounds, making them difficult in pitch, intonation, and musical perception.

The hearing rehabilitation effect of cochlear implants depends on whether the auditory nerve signals caused by electrical stimulation can transmit sound information correctly like the auditory nerve signals of normal people. Studies have found that the cochlea has a frequency analysis function, which transmits acoustic signals of different frequencies to the brain through nerve channels in different parts, and the magnitude of the signal is expressed by the number of pulses in the auditory nerve with the unit of time, that is, the firing rate. This is referred to as the frequency-site, intensity-firing rate coding mechanism for short. The existing cochlear electrons are based on this to formulate corresponding sound signal processing and coding schemes. At present, the voice processing schemes of cochlear implants can be roughly divided into two categories: one is to extract characteristic information such as the fundamental frequency and formant of the voice signal, which is called the SPEAK scheme. For example, the F0/F1/F2 solution used by the Nucleus cochlear implant device; the other type filters the speech signal in frequency bands, and then transmits the corresponding frequency signal to different regions of the cochlea through electrodes according to the cochlear frequency distribution. For example, a continuous interval sampling speech signal processor (Continuous Interleaved Sampling, CIS) is currently widely used. For the above coding schemes, whether it is the SPEAK scheme using the formant or the CIS scheme, the signal processing of the important part of the cochlea is simulated by spectrum analysis or band-pass filtering. They can reflect its frequency-site, intensity-firing rate coding to a certain extent, but cannot reflect the nonlinearity of the cochlea.

We have long been engaged in the research on the signal processing mechanism of the auditory periphery. The latest results of "Laser Interference Research on Cochlear Pitch Information Perception" supported by the National Natural Science Foundation of China show that the cochlea also plays a role in the perception of pitch through nonlinearity, enabling it to The pitch for fundamental-frequency-absent signals and quasi-periodic signals is expressed by nonlinearity. This means that the linear calculations such as spectrum analysis and band-pass filtering used in existing cochlear coding schemes have defects in reflecting cochlear signal processing, and better methods are needed to replace them. The international community has also noticed the importance of the nonlinearity of the cochlea in auditory signal processing. It is believed that the cochlea can be regarded as an oscillator that stays at the Hopf bifurcation point, but it is only a mathematical abstract equation to reflect the cochlea's uniqueness. Some typical non-linear features, there is still a gap between the actual behavior and the real cochlea.

Contents of the invention

Aiming at the defects of the prior art, the purpose of the present invention is to provide a cochlear electronic coding method based on the nonlinear dynamic mechanism of cochlear hearing, aiming at solving the problems of small dynamic range and difficult tone information transmission using the traditional cochlear simulation calculation method. question.

The invention provides a cochlear coding method based on the nonlinear dynamic mechanism of cochlear hearing, which comprises the following steps:

(1) Establish a nonlinear dynamic model of the cochlea;

(2) Construct a nonlinear cochlear array according to the nonlinear dynamic model of the cochlea;

The nonlinear cochlear array is a group of active simulation modules containing n different natural frequencies, and each active simulation module performs corresponding calculations on the received input voice signals according to the cochlear nonlinear dynamics model to obtain each active simulation module. The real-time response output signal of the simulation module;

(3) Rectify and low-pass filter the real-time response output signal of each active simulation module to obtain the amplitude envelope of the real-time response output signal; and convert the frequency and amplitude envelope of the real-time response output signal into a pulse signal Electrodes delivered to different parts;

Wherein, n is the number of active simulation modules, and n is an integer greater than or equal to 1.

Furthermore, the nonlinear dynamic model of the cochlea is: Among them, x is the displacement of the basement membrane from the equilibrium position, t is the time, γ is the damping coefficient, γ _α is the adaptive force coefficient, B is the electrostriction coefficient of the outer hair cell, x0 is the original length of the outer hair cell, and ω _i is the cochlea The natural circular frequency of this part, S(t) is the input voice signal, x _i (t) is the real-time response output signal of the i-th active simulation module, i is the serial number of the active simulation module, i=1,2,3. … n.

Furthermore, the self-adaptive force coefficient γ _α in the nonlinear dynamic model _of the cochlea should satisfy the following range: 0<γ _α ≤ γ. The greater the magnification effect.

Furthermore, the natural frequencies of n active simulation modules can be set as follows: For a nonlinear cochlear array with a natural frequency range of a～a*e ^ε(n-1) Hz (ε<1), the i-th The natural frequency of each active emulation module is f _i =a*e ^ε(i-1) Hz; i is the serial number of the active emulation module, i=1,2,3...n.

Furthermore, 20Hz≤a≤200Hz.

Through the above technical scheme conceived by the present invention, compared with the prior art, due to the introduction of the nonlinear cochlear array using the cochlear nonlinear dynamics model instead of the traditional passive filter bank and the compression module to process the acoustic signal, so that after The processed acoustic signal can exhibit a nonlinear dynamic range compression effect similar to the cochlear processing result, amplifying small signals and compressing large signals, thereby solving the problem of small dynamic range of traditional cochlear electronics. At the same time, the processed acoustic signal can display combined sound information related to the pitch, thereby solving the problems of the traditional cochlear electronic cochlear with a small dynamic range and difficult transfer of pitch information.

Description of drawings

Fig. 1 is a technical block diagram of implementing a novel cochlear electronic speech processing strategy using a cochlear nonlinear dynamic model in the present invention.

Figure 2 is a waveform diagram of the vibration response of the system under the excitation signal of 325Hz+425Hz of the active simulation module.

Fig. 3 is the tuning curve of the active simulation module.

Figure 4 is the image of the response amplitude of the active simulation module changing with the sound intensity, and the response image of the cochlear physiological experiment. Among them, 4(a) is the comparison between the response image of the simulation module and the passive system, and 4(b) is the result of the physiological experiment.

Figure 5 shows the results of nonlinear dynamic compression under different parameters; where Figure 5(a) changes parameter x0 for fixed parameter B, and Figure 5(b) changes parameter B for fixed parameter x0.

Figure 6 shows the two-tone suppression effect in the active emulation module.

Figure 7 shows the response spectrum of the active simulation module and the cochlea response spectrum in the physiological experiment. Figure 7(a) is the response spectrum of the dynamic simulation module, and Figure 7(b) is the response spectrum of the cochlea in the physiological experiment.

Detailed ways

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

The cochlear coding method based on the nonlinear dynamic mechanism of cochlear hearing provided by the present invention is mainly used in the cochlear technology for hearing rehabilitation. Psychoacoustic research has shown that hearing is not a linear perception. For example, when two single-frequency sounds of different frequencies are excited, the listener can not only hear the frequency components of the two excitation signals, but also the frequency components of linear combinations such as their difference frequency (called "combined tone"). Combination tone") exists. When two tones appear at the same time, there will be a two-tone suppression effect in which the two tones sound smaller than when they are separately excited. These nonlinear effects play a very important role in auditory perception. Experimental studies have shown that nonlinear effects such as combined tones, two-tone suppression, and nonlinear dynamic range compression occur during signal processing in the cochlea.

Based on the study of the physiological process of cochlear acoustic signal processing and the understanding of the mechanism of active force generation in hair cells in the cochlea, we established an active dynamics model of cochlear signal processing. This model can not only reflect the nonlinear nature of the cochlea qualitatively, but also reflect its nonlinear dynamic compression and other effects well, and provide results consistent with physiological experimental observations. Based on the results of this basic research, the present invention establishes a cochlear coding method based on the active mechanism of auditory sound perception. By integrating the nonlinearity of the cochlea, it solves the shortcomings of the existing cochlear electronic performance, especially the insufficient auditory dynamic range and the lack of pitch perception.

The concrete technical scheme that the present invention adopts is as follows:

(1) Establishment of the nonlinear dynamic model of the cochlea

Based on the dynamic characteristics of the cochlear basilar membrane, we take the local cochlea as an example to analyze the force. During the process of sound conduction, the cochlear basilar membrane will be subjected to external force F _s (t) caused by external sound stimulation, the basilar membrane's own elasticity F _T =-kx, lymph fluid and its own resistance And the nonlinear self-adaptive force _{F a} regulated by the electrostriction of outer hair cells and the movement of cilia, its simplified expression is as follows: The nonlinear dynamic model of the cochlea established according to Newton's laws of mechanics is as follows: Where x is the displacement of the basement membrane from the equilibrium position, γ is the damping coefficient, γ _α is the adaptive force coefficient, B is the electrostrictive coefficient of the outer hair cell, x0 is the original length of the outer hair cell, and ω _i is the intrinsic circle of the cochlea frequency, the input signal is S(t). The real-time response output x _i (t) of the cochlear basilar membrane excited by the acoustic signal S(t) can be obtained by solving the above nonlinear equation.

(2) Cochlear coding method using the nonlinear dynamic model of the cochlea

As shown in Figure 1, the new cochlear coding method needs to construct a nonlinear cochlear array according to the cochlear nonlinear dynamic model to simulate the cochlea's sound processing mechanism. The nonlinear cochlear array is a group of n designed according to the equation in (1), with different natural frequencies A nonlinear simulation array composed of active simulation modules. The input acoustic signal is S(t), and the real-time response output x _i (t) of different channels after processing can be obtained by numerically solving the above equation. Afterwards, the amplitude envelope is detected by rectification and low-pass filtering, and finally the frequency information and amplitude information of the sound are converted into pulse signals and transmitted to electrodes in different parts. It should be noted that the sharpness of the tuning curve of the simulation module can be changed by adjusting the parameter γ _α during the design of the simulation module. The larger the value of γ _α , the sharper the tuning curve. In general, the adaptive force coefficient γ _α should be set in the range of 0<γ _α ≤γ, when γ _α =0, the adaptive force is zero, and the system becomes a passive system; when γ _α =γ, The system will eventually oscillate on its own. At the same time, the dynamic range and compression range of the simulation module can be changed by adjusting the parameter B or x0 during design, as shown in Figure 5. The cochlear nonlinear dynamic model can well simulate the cochlea's good frequency selection characteristics, dynamic compression characteristics, and multi-tone distortion characteristics, so it replaces the filter bank and compression module in the traditional electronic cochlear sound signal processing scheme to process the sound signal. Processing, a large dynamic response range, and better tonal information transfer results.

The present invention will be described in further detail below in conjunction with the accompanying drawings and specific examples.

Fig. 1 is a technical block diagram of implementing the new cochlear coding by using the nonlinear dynamic model of the cochlea. The specific strategy is as follows: According to the frequency, n frequency band channels need to be designed, and each frequency band channel contains active simulation modules with different natural frequencies, envelope detection modules, pulse modulation modules and electrodes. The acoustic signal recorded by the microphone is preprocessed to obtain S(t), and the output is _xi (t) through different simulation modules, and then _xi (t) is rectified and low-pass filtered to detect its amplitude package Finally, the frequency information and amplitude information of the sound are converted into pulse signals and transmitted to the electrodes.

Figure 2 is the vibration response waveform diagram of the system under the excitation signal of 325Hz+425Hz of the active simulation module; from the spectrum analysis of the waveform in the figure, its frequency components are shown in Figure 7, which shows the combined sound similar to the cochlear physiological experiment observation results .

Figure 3 shows the frequency modulation curve of the simulation module. As shown in the figure, the processing results of the simulation module are similar to the actual cochlear response results, with nonlinear tuning and good frequency selection characteristics. Therefore, it can replace the filter bank in the previous scheme to extract the sound information of different frequency bands.

Figure 4(a) is a graph showing the response of the active system and the passive system as a function of the sound intensity. The dotted line is the response of the passive system, and the solid line is the response of the active system. It can be seen from the figure that the response of the passive system is linear, while the response of the active system has obvious nonlinear characteristics. Due to the linear nature of the passive system response, previous solutions had to use additional compression modules to process the sound in order to obtain greater dynamic range. The compression functions used in traditional schemes are exponential or logarithmic functions. However, the results of physiological experiments (Fig. 4(b)) show that the response characteristic of the auditory system is a nonlinear dynamic compression characteristic, that is, the auditory system can not only realize the compression of large signals, but also the amplification of small signals, which is What cannot be achieved with traditional exponential or logarithmic compression functions. However, the active simulation module can simulate this auditory characteristic very well (Fig. 4(a) solid line), and its dynamic range is close to that of the normal cochlea. Therefore, the active simulation module can replace the compression module in the traditional solution, so that the dynamic range of hearing can be better improved.

Figure 5 shows the effect of nonlinear dynamic compression in the active simulation module under different parameters. It can be seen from Figure 5(a) that as _x0 increases gradually, the curve gradually moves to the positive direction of the x-axis, but the range of the curve compression area does not change with the change of _x0 . The smaller the value of _x0 , the earlier the curve reaches the compressed state. Therefore, if we want to reach the compressed state earlier, we can achieve it by reducing the value of x ₀ . It can be seen from Figure 5(b) that as the parameter B increases, the curve gradually moves to the negative direction of the x-axis, and the range of the curve compression region becomes wider and wider. Therefore, parameter B can be adjusted to obtain a larger compression area.

Figure 6 shows that the active simulation module is able to simulate two-tone suppression in the cochlea.

Figure 7 shows the multi-tone distortion effect. From Figure 7(a), it can be seen that the response spectrum of the simulation module has distortion products that cannot appear in passive systems, namely: combined tones. Fig. 7(b) is the response result on the cochlear basilar membrane in the actual physiological experiment. It can be seen from the comparison that the active simulation module can well simulate the polyphonic distortion effect in the cochlea, which is also the basis for this strategy to improve the ability of the electronic cochlear tone analysis.

The electronic cochlear coding method based on the cochlear nonlinear dynamics mechanism provided by the present invention comprises the following steps:

(1) Establish a nonlinear dynamic model of the cochlea

(2) Construct a nonlinear cochlear array according to the nonlinear dynamic model of the cochlea. The nonlinear cochlear array is a group of simulation arrays containing n active simulation modules with different natural frequencies. computation. The natural frequency of the active simulation module can be set according to the following mode: For example ^, for the nonlinear A cochlear array, wherein the natural frequency of the ith active simulation module is f _i =a*e ^ε(i-1) Hz (i=1,2,3...n). When designing the active simulation module, the self-adaptive force coefficient γ _α should be set in the range of 0<γ _α ≤ γ, the larger the value of γ _α within this range, the greater the amplification effect of the active simulation module on the voice signal near its natural frequency The larger is; the dynamic range and compression range of the simulation module can be changed by adjusting the parameter B or x0 during the design of the active simulation module. The larger the parameter B is, the wider the range of the compression area of the active simulation module will be. The smaller the value of x0, the earlier the curve reaches the compressed state.

The input voice signal is S(t), and the real-time response output x _i (t) of each active simulation module after processing can be obtained by solving the above equation;

(3) The output signal x _i (t) of each active simulation module is rectified, and its amplitude envelope is detected by low-pass filtering. Afterwards, the frequency information and amplitude information of the sound are converted into pulse signals and transmitted to the corresponding electrodes.

Since the active simulation module can realize the dynamic compression of the sound signal, that is, to amplify the small signal and compress the large signal, it replaces the traditional exponential or logarithmic compression function to compress the sound signal, thereby improving the dynamic range of the cochlea. At the same time, the active simulation module can well simulate the polyphonic distortion effect of the cochlea, so the processed signal can show the combined tone related to the tone, so that the electronic cochlear can better transmit the tone information.

It is easy for those skilled in the art to understand that the above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention, All should be included within the protection scope of the present invention.

Claims (5)

1. a kind of cochlear implant coding method based on cochlea sense of hearing Nonlinear Dynamics, which is characterized in that including following Step：

(1) cochlea non-linear dynamic model is established；

(2) non-linear cochlea array is built according to cochlea non-linear dynamic model；

The active emulation module that the non-linear cochlea array includes n different natural frequencies for one group, each is actively emulated After the input speech signal of reception is carried out corresponding computing by module according to the cochlea non-linear dynamic model, obtain each The real-time response output signal of a active emulation module；

(3) the real-time sound is obtained after the real-time response output signal of each active emulation module being carried out rectification and low-pass filtering The amplitude envelope of signal should be exported；And the frequency of real-time response output signal and amplitude envelope are converted into pulse signal and passed Pass the electrode of different parts；

Wherein, n is the number of active emulation module, and n takes the integer more than or equal to 1.

2. cochlear implant coding method as described in claim 1, which is characterized in that the cochlea non-linear dynamic model For：

Wherein, x deviates equilbrium position displacement for basilar memebrane, and t is the time, and γ is damped coefficient, γ_αFor adaptive force coefficient, B is outer Hair cell electrostriction coefficient, x0 is former long for external hair cell, ω_iFor the inherent circular frequency at the cochlea position, S (t) is input language Sound signal, x_i(t) real-time response for i-th active emulation module exports signal, and i is the sequence number of active emulation module, i=1, 2,3......n。

3. cochlear implant coding method as claimed in claim 2, which is characterized in that in the cochlea non-linear dynamic model Adaptive force coefficient γ_αFollowing scope should be met：0 ＜ γ_α≤ γ, within this range γ_αValue is bigger, and active emulation module is solid to it The amplification for having the voice signal near frequency is bigger.

4. such as claim 1-3 any one of them cochlear implant coding methods, which is characterized in that n active emulation module Intrinsic frequency can be set as follows：

It is a~a*e for intrinsic frequency scope^ε(n-1)The non-linear cochlea array of Hz (ε ＜ 1), wherein i-th actively emulates mould The intrinsic frequency of block is f_i=a*e^ε(i-1)Hz；I be active emulation module sequence number, i=1,2,3......n.

5. cochlear implant coding method as claimed in claim 4, which is characterized in that 20Hz≤a≤200Hz.…….