【作者】 杨丹； 徐彬； 王旭；

【Author】 Yang Dan~1 Xu Bin~2 Wang Xu~1 (1 School of Information Science & Engineering Northeastern University Shenyang 110004 China; 2 Computing Center of Northeastern University Shenyang 110004 China)

【机构】 东北大学信息科学与工程学院； 东北大学计算中心；

【摘要】 声音在作用时间不同时,听觉感受性将起变化,提示了听觉中枢以一定的时间单元对传入信息整合,信息的累积有利于分辨率的锐化。为了更好地看到语音信号的"精细结构",提出了一种适合听觉特性的语谱分析方法。该方法根据傅立叶变换原理,结合听觉特点,逐点计算信号频率,并通过逐次累加适当衰减得到瞬时频谱。实验表明,当衰减系数选取适宜时,算法能够很好地模拟听觉时间特性,提高语音频谱的分辨率,克服测不准定理对时间分辨率和频率分辨率的限制,是对传统语谱分析算法的提升。更多还原

【Abstract】 Auditory perception will change with different durations of the sound,suggesting that the hearing system summates or integrates sound over time unit and the sharpening of resolution needs summation of afferent information.In order to find the"fine structure"of speech signal,it is proposed that a new method of speech spectrogram using auditory characteristics.The method analyses the frequency components of speech signal point by point based on Fourier Transform,followed by summation and attenuation and then achieve the instant spectrum.It shows that when the attenuation parameter is appropriate,the method will simulate the characteristic of hearing system well and enhance the resolution of speech spectrogram without the limitation of time and frequency resolution based on uncertainty principle.It made an improvement for the traditional speech spectrogram computation.更多还原