Study On The Generation Method Of Auditory Spectrum Based On Auditory Bionics

In recent years, with the development of biological sciences and physiology, theresearch of human organs is more thoroughly. After ten million years of evolution, thehuman auditory system has the incomparable performance in sound recognition area.Research shows that the auditory system ultimately represents sound characteristicswith auditory spectrum; therefore, it is significant to study on the generating methodof auditory spectrum based on auditory bionics both in sound recognition and auditorybionics area. This paper complete the study mainly from the following three aspects:First, The work mechanism of the auditory system, including the basilarmembrane, inner hair cells, auditory nerve, cochlear nuclei and inferior colliculus arestudied through analyzing the physiological structure of them. Thus enoughphysiological basis for the generation of auditory spectrum is acquired.Second, Based on the study of the previous model and the physiological data, arelatively complete auditory system model is set up. the frequency selectivecharacteristics of the basilar membrane, the dynamic transfer of Neurotransmitterbetween haircell and nerve fiber are successfully modelled. Then a lateral inhibitionnetwork to model the lateral inhibition character between the cochlear nuclei neuronsis established. a short time integral to model the lost of phase-locking accrued in thecentral auditory system neurons is used.Third, the auditory system models are simulated in Matlab. Simulation resultshows that the Gammatone Filter bank has excellent frequency selection characteristic.The output of the Inner Haircell-Auditory Nerve model has perfect ‘onset’ andphase-locking characteristic for a sinusoidal input signal, and it fit the physiologicaldata very well. At the same time, the outputs of the lateral inhibition network showexcellent lateral inhibition characteristic that exists between the nuclei neurons.Finally，75segments of voice signal for5kinds of instruments are imported tothe auditory system model, and75auditory spectrums are created. The acquiredspecturms show that auditory spectrums can be used as good temporal and spectralfeatures feathers for different musical instruments. This method can be further used ininstrument recognition and timbre discrimination.