| 基于FSS与PLP的噪声鲁棒语音识别 |
| |
| 引用本文： | 王振力,白志强,朱江.基于FSS与PLP的噪声鲁棒语音识别[J].南京邮电大学学报(自然科学版),2008,28(4). |
| |
| 作者姓名： | 王振力 白志强 朱江 |
| |
| 作者单位： | 1. 南京国际关系学院,博士后流动站,江苏,南京,210039
2. 南京陆军指挥学院,军事运筹教研室,江苏,南京,210045 |
| |
| 基金项目： | 中国博士后科学基金，江苏省博士后科学基金 |
| |
| 摘 要： | 提出了一种基于分数阶谱相减(FSS)与感知线性预测(PLP)相结合的噪声鲁棒语音识别方法,记为FSS PLPC.该方法首先通过FSS在分数阶Fourier域对带噪语音进行降噪处理,然后计算增强语音的均方误差和Itakura距离并进行比较,以获得FSS的近似最优分数阶阶数.最后对根据此阶数得到的增强语音提取感知线性预测倒谱(PLPC).实验结果表明,FSS PLPC对于数字语音的识别性能优于传统的谱减法(SS PLPC)和感知线性预测倒谱(PLPC)法,并且随着信噪比的降低FSS PLPC表现出较好的噪声鲁棒性.
|
| 关 键 词： | 噪声鲁棒语音识别 语音增强 谱减法 分数阶Fourier变换 感知线性预测 |
Noise Robust Speech Recognition Based on Fractional Spectral Subtraction and Perceptual Linear Prediction |
| |
| WANG Zhen-li,BAI Zhi-qiang,ZHU Jiang.Noise Robust Speech Recognition Based on Fractional Spectral Subtraction and Perceptual Linear Prediction[J].Journal of Nanjing University of Posts and Telecommunications,2008,28(4). |
| |
| Authors: | WANG Zhen-li BAI Zhi-qiang ZHU Jiang |
| |
| Institution: | WANG Zhen-li1,BAI Zhi-qiang1,ZHU Jiang21.Postdoctoral Station,Nanjing University of International Relations,Nanjing 210039,China2.Department of Military Operational,Nanjing Army Comm, College,Nanjing 210045,China |
| |
| Abstract: | In this paper,a method named FSS+PLPC by combining Fractional Spectral Subtraction(FSS) and Perceptual Linear Prediction(PLP) is proposed for noise robust speech recognition.It applies FSS to noisy speech to reduce noise components in the fractional Fourier domain.An approximate optimal fractional order is then obtained by computing Mean Square Error(MSE) and Itakura distance of the enhanced speech and comparing the difference between them.Finally,Perceptual Linear Predictive Cepstral(PLPC) is computed acco... |
| |
| Keywords: | Noise robust speech recognition Speech enhancement Spectral Subtraction FRactional fourier transform Perceptual linear predictive |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
