论文部分内容阅读
提出主元分析 PCA(Principal Component Analysis)用于语音检测的方法研究.用主元分析法在多维空间中建立坐标轴,将待处理信号投影到该坐标轴中,通过分析投影结果判断是否为语音信号.通过将语音和非语音分别建立子空间,来区分语音和非语音信号.该方法不同于常规的语音时域、频域处理方法,而是在多维空间中对信号进行分析.实验结果表明,该方法准确率高、简单、容易实现,而且能区分多种非语音信号.
A Principal Component Analysis (PCA) method for speech detection is proposed. Principal component analysis (PCA) is used to establish coordinate axes in multi-dimensional space, and the signals to be processed are projected onto the coordinate axes. Signal.This method distinguishes speech and non-speech signals by establishing subspace separately for speech and non-speech.This method is different from the conventional speech time-domain and frequency-domain processing methods, but analyzes signals in multi-dimensional space.The experimental results show that The method is accurate, simple and easy to implement, and can distinguish a variety of non-speech signals.