论文部分内容阅读
提出了一种采用感知语谱结构边界参数(PSSB)的语音端点检测算法,用于在低信噪比环境下的语音信号预处理。在对含噪语音进行基于听觉感知特性的语音增强之后,针对语音信号的连续分布特性与残留噪声的随机分布特性之间的不同点,对增强后语音的时-频语谱进行二维增强,从而进一步突出连续分布的纯净语音的语谱结构。通过对增强后语音语谱结构的二维边界检测,提出PSSB参数,并用于端点检测。实验结果表明,在白噪声-10 dB到10 dB的各种信噪比环境下,采用PSSB参数的端点检测算法,相对于其它端点检测算法,更有效地检测出语音的端点。在-10 dB的极低信噪比下,提出的方法仍然有75.2%的正确率。采用PSSB参数的端点检测算法,更适合于低信噪比白噪声环境下的语音端点检测。
A speech endpoint detection algorithm based on PSSB is proposed for speech signal preprocessing in low SNR environment. After speech enhancement based on auditory perception of noisy speech, the time-frequency spectrum of enhanced speech is two-dimensionally enhanced according to the difference between the continuous distribution of speech signal and the random distribution of residual noise. Thereby further highlighting the continuous distribution of pure speech spectrum structure. Through the two-dimensional boundary detection of the enhanced phonetic structure, PSSB parameters are proposed and used for endpoint detection. The experimental results show that the endpoint detection algorithm based on PSSB parameters can detect the endpoint of speech more effectively than other endpoint detection algorithms in various signal-noise environments with white noise of -10 dB to 10 dB. At a very low signal-to-noise ratio of -10 dB, the proposed method still has a 75.2% correct rate. The endpoint detection algorithm using PSSB parameters is more suitable for voice endpoint detection under low signal-to-noise white noise environment.