摘要
目的研究不同强度白噪声对语音特征提取的影响,总结变化规律,为带噪语音的声纹鉴定提供参考。方法对TIMIT连续语音语料库中的录音样本加载不同强度白噪声,使用语音工作站提取纯净语音、不同信噪比带噪语音的基频、共振峰,研究分析白噪声对语音特征参数的影响。结果低噪环境下,语音共振峰相对稳定,增大噪声强度,共振峰出现偏移或者无法检出现象;各阶共振峰抗噪能力不同,低阶共振峰抗噪声能力强,稳定度高,高阶共振峰抗噪能力弱,稳定性差,低阶共振峰抗噪能力优于高阶共振峰;基音在各种噪声强度下稳定度高,具备较强鲁棒性。结论信噪比降低会引起共振峰频率偏移,甚至丢失共振峰;噪声对高阶共振峰影响大于低阶共振峰;基频在噪声环境下具备较高的抗干扰能力,声纹鉴定中应重点分析噪声对语音特征的影响。
Objective To study the influence of white noise on the extraction of phonetic feature and to offer a reference for identification of the speech mixed with noises.Methods After embedding the different intensity of white noise into the speech samples which were selected from the DARPA TIMIT acoustic-phonetic continuous speech corpus,we utilized speech workstation to extract pitches and formants from both pure speech and noisy speech of different signal-to-noise ratio(SNR) thereby analyzing the influence of white noise on the parameters of phonetic feature.Results Formant frequency kept relatively stable in low noisy condition but appeared shifting or even undetected under certain conditions with the noise strength enhancing.For each order of formant,the anti-noise ability was different,with the low order showing strong and highly stable but the high order weak and poorly stable.However,the pitch retained steady in various noisy conditions.Conclusions Reducing SNR can result in formant frequency shifting or vanishing.The influence of noise on the formant of high order is bigger than that of the low order.Pitch keeps highly stable in noisy environment.
出处
《刑事技术》
2015年第1期49-54,共6页
Forensic Science and Technology
基金
智能语音技术公安部重点实验室开放课题(No.2014ISTKFKT02)
关键词
声纹
线性预测系数
白噪声
共振峰
基音
voiceprint
linear prediction coefficient
white noise
formant
pitch