话音检测器和话音瞬态参数的统计分析

本文介绍一种用于话音内插系统(Speech Interpolation System)的高灵敏度话音检测器,并用该检测器进行了各种话音瞬态参数的统计分析。该话音检测器通过分析输入信号的短时能量和零交叉率等检测话音信号,采用自适应电平门限和释放延迟等措施提高了话音检测器的灵敏度和可靠性。同时,这种方案对于线路中的信令信息有较强的识别能力。采用本话音检测器对话音信号所做的统计分析不仅得到了具有一定价值的汉语活动规律,而且进一步完善了检测方案。同时,通过汉语和英语话音活动规律的比较表明汉语应用话音内插技术的效果要优于英语。

Speech Detector and Statistical Analysis of Speech Temporal Parameters

This paper presents a sensitive speech detecter used in Speech Interpola- tion Systems and deals with the measurement of various speech temporal parameters. The speech detector detects the presence of speech signals by analysing short-time energy and zero-cross rate of input signals.The utilization of a self-adaptive energy threshold and a short time hangover reduces the speech clipping and makes the detctor more reliable and more flexible. This detector can also identify the long-distance signalings. The statistical analysis of the speech temporal parameters not only illustrates the significant on-line speech activity, but also perfects the detective method. The comparison of the speech temporal parameters between Chinese and English speech occasions shows that the SI technique is more suita- ble to Chinese occasion than English occasion.