Psychophysical Study On Human Auditory Information Processing Mechanism

Sound signal reaching human ears is usually complex. It is composed of severalcomponents coming from different sound sources. However, human auditory systemusually can extract and trace the component which the listener is interested in. It is such aremarkable ability that no any machine can be compared. Study of the human auditoryinformation processing mechanism is one important part in the the human brain cognitiveresearch field. Moreover, it has important reference values to the automatic speechrecognition and automatic speaker recognition techniques. Based on the psychophysicallevel, we studied the following three aspects of the human auditory information processingmechanism.In the complex auditory environment, the sound signal reaching ears is a mixture oftarget sound and masking sound. Previous research has demonstrated that the listener’sprior knowledge of the target speech cues can help the listener to extract the target fromthe masking background. However, less data are avaliable concerning the influence of themasker cues on the ability to recognize the target speech. A psychoacoustic experimentconducted in the first work of this thesis indicates that cueing the listener to the masker fora short time actually impairs, rather than improves, the performance in the target speechrecognition task. It suggests that the short-term prior knowledge of the masker enhancesthe informational masking and the listener’s attention is partially drawn to the masker,which make the target recognition difficult. We conclude that the human auditory systemis involved in a performance of only enhancing the familiar target, but not suppressing thefamiliar masking speech signals.Auditory signals received by human ears all have statistical characters. Differentphonemes in the speech signals have different probabilities of occurrence in naturalauditory enviorment. In the second work of this thesis, a speaker recognitionpsychoacoustic experiment was conducted to study the human auditory perception andprocessing mechanism to the different Chinese vowels during the speaker recognitionprocess. The experiment result shows that the subjects responded more to the low probability vowels than to the high probability vowels, which indicates that the lowprobability vowels conveyed more personal voice information than the high probabilityvowels. This work also illustrates that the human auditory system performs non-uniformweight processing of the vowel signals during the speaker recognition process, in whichthe low probability vowels are finely encoded, while the high probability vowels areroughly encoded, so that the human auditory neural resources can be efficiently utilizedand the voice information transmission can be maximized. This work provides a humancognitive behavioral evidence for the application of the information theory and theefficient coding theory in the human auditory system.Human auditory system has acute perception ability to auditory signals. It can finelydistinguish the interaural time difference which is in microseconds. While the auditorysignal is relatively coarse because its time resolution is higher than milliseconds and thetemporal pattern is uncertain. How the uncertain and coarse auditory signal can transferthe fine temporal information has long been an intriguing puzzle. In the third work of thisthesis, we demonstrate that the neural signal, which is uncertain due to noise, can besynchronized by noise at the same time by the numerical calculation on theAdelman-FitzHugh neural model. This uncertainty and synchrony of neural signal inducedby noise can be used to explain the auditory capability of transferring the fine temporalinformation.Above work provides some useful human cognitive experimental evidence andreference theoretical explanations to the human auditory information processingmechanism.