压缩感知技术在语音增强和EEG信号中的应用研究

发布时间：2018-07-17 06:46

【摘要】：语音信号和脑电(EEG)信号是数字信号处理的重要内容,具有广泛的应用背景。实际中,这些信号通常是非稳定并夹杂噪声。研究从带噪的信号中提取纯净信号能提高应用效果,不仅具有重要的学术价值,而且具有应用意义。本文的主要工作包括两部分：第一部分,通过研究压缩感知技术和语音增强算法,提出了一种基于压缩感论技术的语音增强新算法。新算法使用的技术是采用KLT (Karhunen-Loeve Transform)矩阵作为变换矩阵,通过分析纯净语音信号与噪声信号在KLT矩阵下稀疏性的差异,将带噪语音信号在KLT矩阵上进行系数投影,利用正交匹配追踪(Orthogonal Matching Pursuit,OMP)算法获得稀疏系数并对该稀疏系数进行线性过滤修正,最后进行信号重构。实验结果证实在各种白噪声与有色噪声环境下,本文提出的基于压缩感知技术的语音增强算法优于经典的语音增强算法。第二部分,研究了EEG信号处理方法及传统的白噪声去噪预处理方法,将提出的基于压缩感知技术的语音增强新算法应用于EEG信号去噪预处理,进而提出处理EEG信号的改进算法。改进算法的特征是对预先采集的噪声进行估计获得自相关矩阵,应用上述提出的语音增强新算法对EEG信号进行滤波实验结果证明,本文提出的改进方法对EEG信号的去噪效果和去噪后ERP(事件相关电位)保留都优于基于NLM (Nonlocal means)的EEG信号处理方法。
[Abstract]:Speech signal and EEG signal are important contents of digital signal processing, and have wide application background. In practice, these signals are usually unstable and noisy. The study of extracting pure signals from noisy signals can improve the application effect, which is not only of great academic value, but also of practical significance. The main work of this paper consists of two parts: in the first part, a new speech enhancement algorithm based on the theory of compression perception is proposed by studying the compression sensing technology and the speech enhancement algorithm. The new algorithm uses the KLT (Karhunen-Loeve transform) matrix as the transformation matrix. By analyzing the difference between the sparsity of the pure speech signal and the noise signal under the KLT matrix, the noisy speech signal is projected on the KLT matrix. The sparse coefficient is obtained by orthogonal matching pursuit (OMP) algorithm, and the sparse coefficient is modified by linear filtering. Finally, the signal is reconstructed. The experimental results show that the proposed speech enhancement algorithm based on compression sensing is superior to the classical speech enhancement algorithm under various white and colored noise environments. In the second part, the EEG signal processing method and the traditional white noise denoising preprocessing method are studied. A new speech enhancement algorithm based on compression perception is applied to the EEG signal denoising preprocessing, and an improved EEG signal processing algorithm is proposed. The feature of the improved algorithm is to estimate the pre-acquired noise to obtain the autocorrelation matrix. The experimental results show that the proposed new speech enhancement algorithm is used to filter EEG signals. The improved method proposed in this paper is superior to the NLM-based non-local means method for the denoising effect and the event-related potential (ERP) retention after denoising.
【学位授予单位】：福州大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN912.35

【相似文献】