引用本文
  •    [点击复制]
  •    [点击复制]
【打印本页】 【下载PDF全文】 查看/发表评论下载PDF阅读器关闭

←前一篇|后一篇→

过刊浏览    高级检索

本文已被:浏览 368次   下载 856 本文二维码信息
码上扫一扫!
高噪声环境下的语音活动检测技术
0
()
摘要:
以MOTOROLA公司生产的DSP56F801芯片为核心搭建系统硬件电路,对语音信号进行处理。采用能够根据背景噪声变化而自适应调节门限的算法对语音活动进行精确检测,实现整机系统在高噪声环境下的正常通讯。实际检测结果表明:在100、110、115dB的强背景噪声环境下,系统识别率分别为100%、96%和94%,丢字漏字率分别为2%、5%和10%;在100~110、110-115dB的背景噪声环境,系统稳定性分别为100%和98%。采用此种算法,增强了系统的抗干扰能力,在高噪声和噪声不太稳定的环境下,实现了对语音活动的精确检测。
关键词:  语音活动检测 短时能量 过零率 短时自相关 自适应门限
DOI:10.11841/j.issn.1007-4333.2005.03.072
修订日期:2004-11-25
基金项目:
Voice detection technology (VDT) in a background of high noise
Abstract:
In order to identify the voice activity accurately in a background of 100-115dB high noise, voice signals were interpreted by using a circuit based on the DSPF801 chip produced by the Motorola. The voice detection algorithm used in the circuit could adjust threshold automatically according to the change of noise signals, which ensured a normal communication of the whole system. The test results showed that the detection rate reached 100%, 96% and 94% corresponding the noise levels of 100, 110 and 115dB respectively; and the words loss rate were 2%, 5% and 10% respectively. The system stability reached 100% and 98% respectively in high noise background of 100-110dB and 110-115dB. The system ability on anti-jamming was improved and a precision detection of VD was implemented in a high noise background or to unstable noise signals by the developed algorithm.
Key words:  VAD (voice activity detection),short-time energy,zero-cross ratio,short-time auto-correlation,adaptive threshold