改进的快速独立分量分析在语音分离系统中的应用
|
摘要
为了提高语音分离系统的处理速度和去除其内部噪声,提出了改进的快速独立分量分析算法。首先,运用时频二值掩蔽模型,在前期的处理中滤除语音数据中的部分噪声。其次,运用牛顿插值的思想,用一阶均差来代替迭代中的求导,提高语音信号分离速度。最后,建立系统分离模型和三麦克风阵列,并进行改进算法与经典算法的处理时间的对比实验。实验结果表明,该算法可以实现对一段包括人声、音乐声和随机噪声的混合语音的分离,并且处理效率比经典算法提高了13.5%。
In order to improve the processing speed and remove internal noise in speech separation systems, a modified Fast Independent Component Analysis algorithm(FastICA) was proposed. Firstly, the time-frequency binary masking model was used to filter out part of noise in the speech data in the pre-processing stage. Then, the derivation in iteration was replaced by the first-order mean difference based on Newton interpolation, improving the speech signal separation speed. Finally, a system separation model and a three-microphone array were established, and the comparison experiment of processing time between the improved algorithm and the classical algorithm was carried out. The experimental results show that the proposed algorithm can realize the separation of a mixed speech including voice, music and random noise, and has 13.5% improvement in processing efficiency compared to the classical algorithm.
In order to improve the processing speed and remove internal noise in speech separation systems, a modified Fast Independent Component Analysis algorithm(FastICA) was proposed. Firstly, the time-frequency binary masking model was used to filter out part of noise in the speech data in the pre-processing stage. Then, the derivation in iteration was replaced by the first-order mean difference based on Newton interpolation, improving the speech signal separation speed. Finally, a system separation model and a three-microphone array were established, and the comparison experiment of processing time between the improved algorithm and the classical algorithm was carried out. The experimental results show that the proposed algorithm can realize the separation of a mixed speech including voice, music and random noise, and has 13.5% improvement in processing efficiency compared to the classical algorithm.
引文
[1] 谷岳.移动机器人听觉导航系统中语音分离技术研究[D].天津:河北工业大学,2011:6-10.
[2] 王菁.基于计算听觉场景分析的混合语音分离[D].秦皇岛:燕山大学,2016:8-15
[3] 张政.基于独立分量分析的盲源分离算法优化研究[D].南京:南京信息工程大学,2015:2-10
[4] 黄艳波.单通道盲源分离算法的研究[D].杭州:杭州电子科技大学,2017:11-15.
[5] 粟嘉,陶海红,饶烜,等.时频面滑窗掩膜的多分量信号高效重构算法[J].电子与信息学报,2015,37(4):804-810.
[6] 钱思冲,向阳,李胜杨,等.基于独立分量分析与二值掩膜的语音分离[J].华中科技大学学报(自然科学版),2015,43(7):87-92.
[7] 高波.欠定盲源分离及其应用[D].大连:大连理工大学,2009:1-56.
[8] 朱忠奎,陈祥芹,樊薇,等.基于降维的独立分量分析及信号特征成分检测[J].仪器仪表学报,2014,35(4):917-923.
[9] RADFAR R H,DANSEREAU R M,CHAN W-Y.Monaural speech separation based on gain adapted minimum mean square error estimation[J].Journal of Signal Processing Systems,2010,61(1):21-37.
[10] PERTI? P,NIKUNEN J.Distant speech separation using predicted time-frequency masks from spatial features[J].Speech Communication,2015,68(6):97-106.
[11] 张杰.FastICA算法研究及其在语音信号分离中的应用[D].长沙:湖南师范大学,2013:1-73.
[12] ZHANG Y,ZHAO Y.Modulation domain blind speech separation in noisy environments[J].Speech Communication,2013,55(10):1081-1099.
[13] 王刚.基于最大非高斯估计的独立分量分析理论研究[D].长沙:国防科学技术大学,2005:13-42.
[14] CHEN J,WANG Y,WANG D.Noise perturbation for supervised speech separation[J].Speech Communication,2016,78(4):1-10.
[15] 刘少华.基于独立分量分析的语音信号分离及在数据采集卡上的实现[D].桂林:广西师范大学,2013:30-36.
[2] 王菁.基于计算听觉场景分析的混合语音分离[D].秦皇岛:燕山大学,2016:8-15
[3] 张政.基于独立分量分析的盲源分离算法优化研究[D].南京:南京信息工程大学,2015:2-10
[4] 黄艳波.单通道盲源分离算法的研究[D].杭州:杭州电子科技大学,2017:11-15.
[5] 粟嘉,陶海红,饶烜,等.时频面滑窗掩膜的多分量信号高效重构算法[J].电子与信息学报,2015,37(4):804-810.
[6] 钱思冲,向阳,李胜杨,等.基于独立分量分析与二值掩膜的语音分离[J].华中科技大学学报(自然科学版),2015,43(7):87-92.
[7] 高波.欠定盲源分离及其应用[D].大连:大连理工大学,2009:1-56.
[8] 朱忠奎,陈祥芹,樊薇,等.基于降维的独立分量分析及信号特征成分检测[J].仪器仪表学报,2014,35(4):917-923.
[9] RADFAR R H,DANSEREAU R M,CHAN W-Y.Monaural speech separation based on gain adapted minimum mean square error estimation[J].Journal of Signal Processing Systems,2010,61(1):21-37.
[10] PERTI? P,NIKUNEN J.Distant speech separation using predicted time-frequency masks from spatial features[J].Speech Communication,2015,68(6):97-106.
[11] 张杰.FastICA算法研究及其在语音信号分离中的应用[D].长沙:湖南师范大学,2013:1-73.
[12] ZHANG Y,ZHAO Y.Modulation domain blind speech separation in noisy environments[J].Speech Communication,2013,55(10):1081-1099.
[13] 王刚.基于最大非高斯估计的独立分量分析理论研究[D].长沙:国防科学技术大学,2005:13-42.
[14] CHEN J,WANG Y,WANG D.Noise perturbation for supervised speech separation[J].Speech Communication,2016,78(4):1-10.
[15] 刘少华.基于独立分量分析的语音信号分离及在数据采集卡上的实现[D].桂林:广西师范大学,2013:30-36.