Research On The Auditory Information Processing Method Based On Vibration Theroy And Neuron Filtering Mechanism

Auditory information processing is an essential process for living things to perceive external information, furthermore, is one of significant ways for information exchange between living things and their surroundings. It has the the property of interdisciplinary. With the developments of bionic science, information technology, etc, the achievements of auditory information processing have wide applications in areas such as robot’s interactive processing, information perception and identification, and navigation and positioning, etc. Currently, auditory modeling is the main method for researching on auditory information processing. Researchers study on the process of auditory information processing by modeling auditory receptor, and obtain the results which are closer to auditory characteristics. However, the model of auditory receptor is limited in auditory biomimetic applications for its massive infrastructure, complitated derivation and calculation process. This forced researchers directly focus on the kernel mechanism of the auditory information processing, simulate the function of basilar membrane in cochlea, and develop the model of basilar membrane. It implements simply, but ignores the functions of other structures in auditory receptor. Consequently, it has some shortcomings in terms of auditory response consistency. In response to the problems mentioned, in order to obtain an auditory infromation method which implements simplely and obtain the results which is closer to auditory characteristics, researchers continuously look for the novel approach from perception and processing mechanism of living things.We analyse the auditory mechanism elaborately, and focus on the mechanism of neuron filtering mechanism, generalize the process that external information processed by living things into the process of “Receptor perception—Neuron filtering”. On that basis, we research on the “Auditory receptor—Neuron”, imulate the auditory information processing process integrally from macro information flow and micro information processing mechanism, sand convert the information processing in auditory receptor into the neuron response of auditory information by the keep information mechanism of auditory nerves. We focus on the information transmission form in auditory receptor and neuron, propose an auditory receptor model and a neuron filtering model by vibration theory, develop an auditory information processing method which realize better auditory response consistency and has a simple function form. Specific studies include the following aspects.By analyzing the mechanism of auditory information processing and considering the physiological stucture of auditory receptor, we set up the three-dimensional stucture of cochlear perceptual model. By means of vibration equivalent analysis and conformal mapping, we simulate the vibration of basilar membrane in cochlear internal fluid environment and the coupling function between basilar membrane and cochlear fluid. By cross-section method, we study on the microcosmic mechanics characteristics in cochlea, including the active force of outer hair cells and the coupling between basilar membrane and tectorial membrane. On that basis, we develop the cochlear perceptual model based on vibration theory and auditory mechanism. By means of the simulation experiments and the comparison with auditory physiology data, we verify the position characteristics of frequency distribution, the nonlinear amplication function, and the high auditory acuity for the cochlear perceptual model based on vibration theory and auditory mechanism.We analyse the feasibility of simulating the neural information transmission process by vibration theory. Analyse the neuron filtering mechanism elaborately. By improving the FHN neuron model, we set up a neuron model based on incremental change of membrane potential from the point of vibration. Introduce the neuron filtering mechanism into the neuron model based on incremental change of membrane potential, and propose the neuron filtering model based on vibration theory. Study on the characteristics of parameters, the stable parameter condition and the parametric selection strategy respectively. By means of simulation experiments, tests via hardware and the comparative analysis, we verify the effectiveness, the applicability, the objective denoising performance and the real-time of the neuron filtering model based on vibration theory.By analysing the information flow in “Auditory receptor—neuron” and considering the cochlear perceptual model and the neuron filtering model developed in this paper, we study on the relationship conditions between the cochlear perceptual information and the corresponding response of neuron filtering to realize the connection of cochlear perceptual model and neuron filtering model. On that basis, we develop an auditory information processsing method based on cochlear perception and neuron filtering. Study on parameters’ stable condition, selection strategy, calculating equations, and determinate process respectively. By means of simulation experiments, we study and verify the auditory response consistency of proposed method from three aspects including nonlinear amplification, frequency-domain response, and time-domain response. Comparing with the representative establishment methods of basilar membrane model, we evaluate the performances of our proposed method.We study on the experiments of our proposed cochlear perceptual model based on vibration theory and auditory mechanism for simultaneous perception of multiple voices. The cochlear auditory characteristics are analyzed by experiments of auditory masking. Then we experiment on the proposed auditory information processing method, and respectively introduce the method to sound stimulus, voice reproduction, sound enhancement, speech feature extraction and speaker recognition. We obtain the responses of pure tones and complex tones which satisfy the frequency analysis characteristics of cochlea, realize the reproduction of voice frequency spectrum and obtain the frequency spectrums which satisfy the function of cochlear nonlinear amplification, enhance the objective evaluation index of denoising ability and the subjective loudness simultaneously, extract the speech feature parameters which show more efficient to speaker recognition and much sharper to the high frequency information in voice, show better recognition ratio via experiments and comparisons in speaker recognition under noise free condition or noise condition.