基于ICA的语音信号盲分离.docVIP

基于ICA的语音信号盲分离  [摘 要]语音信号盲分离处理的含义是指利用BSS技术对一段语音信号进行处理。混合语音信号的分离是盲分离的重要内容，目前的混和语音分离大多是建立在无噪环境中的混叠情形下，主要以盲源分离(Blind Source Separation，BSS)，根据信号的统计特性从几个观测信号中恢复出未知的独立源成分。 本文重点研究了以语音信号为背景的盲源分离，在语音和听觉信号处理领域中，如何从混有噪声的的混叠语音信号中分离出各个语音源信号，来模仿人类的语音分离能力，成为一个重要的研究问题。具体实现主要结合ICA技术，将语音去噪作为一个预处理过程，对带噪声的混叠语音盲分离进行了研究，本文详细了介绍了FastICA算法，将这种算法应用于实际的语音信号噪声分离中，并将分离出的语音信号与混合前的原信号进行了分析比较，验证了通过ICA实现语音??号的盲分离是切实可行的。 [关键词]语音信号，盲源分离，独立成分 [Abstract]Blind separation of speech signals processing means for processing refers to a section of the speech signal of microphone detected by BSS technique. Separate the mixed speech signals is an important content of blind separation, the mixed speech separation is mostly based on noise free environment in the overlapping case, mainly to the blind source separation (Blind Source Separation, BSS), according to the statistical characteristics of the signal from the observed signals recover independent source component is unknown. This paper focuses on the blind source separation using speech signal as the background, in speech and audio signal processing field, separating each voice source signal from noisy mixed speech signals, to mimic human speech separation ability, has become an important research question. Concrete realization mainly with ICA technology, the speech denoising as a pretreatment process, the overlapping speech blind separation of mixed with noise is studied, this paper presents the FastICA algorithm, the speech signal noise separation of this algorithm is applied to the practice, and the original voice signal mixed with isolated anterior are analyzed and compared, verified by ICA to realize the blind separation of speech signals is feasible. [Keywords]speech signal, blind source separation, independent component analysis 目录  TOC \o 1-3 \f \h \z \u  HYPERLINK \l _Toc357710791 1 前 言  PAGEREF _Toc357710791 \h 1  HYPERLINK \l _Toc357710792 1.1 盲语音信号分离技术的背景及意义  PAGEREF _Toc357710792 \h 1  HYPERLINK \l _Toc357710793 1.2 语