Precise Molecular Model of Human Cochlea System and Its Application to Speech Recognition

本研究的目的是基于人类外周听觉系统的数学模型来实现高性能的语音信号处理。为了验证该模型的有效性，对该模型产生的多维脉搏信号进行了分析。结果表明，该外周听觉模型具有频率选择性、双音抑制、掩蔽等人类听觉特性。并将该模型应用于噪声环境下的说话人辨认。实验结果表明，在噪声水平较高的情况下，基于周边听觉模型的方法比传统方法具有更好的性能。

The purpose of this study is to achieve a high performance speech signal processing based on a mathematical model of the human peripheral auditory system. In order to verify the validity of the model, the multi-dimensional pulse signals generated by the model are analyzed. As a result, it is confirmed that the peripheral auditory model has the characteristics of the human auditory sense such as frequency selectivity, two-tone suppression, masking, and so on. Furthermore, the model is applied to speaker identification in noisy environment. Experimental results show that the proposed method based on the peripheral auditory model has a better performance than the conventional methods in case where the noise level is high.