论文部分内容阅读
文章提出了一种新的语音信息检测的较灵活的方法。其中用到了两种技术:人工神经网络和复倒谱矩阵。目的是如果用人工神经网络就能够较彻底地解决未明确定义的映射关系。对各种在较低的噪音信噪比值情况下观察结果都有较高的可信度。在语音信号检测过程中,由于语音的特征文章利用线性预测系数得到复倒谱矩阵,这样做会以最低的代价提供较高的对数频谱的估计程度,并且提高了频谱域和时域的有效性。文章测试了几种不同的W SS噪声以及不同信噪比(SNR)的情形,在3dB～10dB的范围之内,AN N方法显著地优于利用语音信号的能量和过零率检测的方法,同时也提高了其它基于复倒谱矩阵方法的准确率。
The article presents a new method of voice information detection more flexible. Two techniques are used: artificial neural network and complex cepstrum matrix. The purpose is to use artificial neural networks can be more thorough solution to the mapping is not clearly defined. For all kinds of low noise signal to noise ratio in the case of the observations have a higher credibility. In the voice signal detection process, since the speech feature articles use the linear prediction coefficients to obtain the complex cepstrum matrix, doing so will provide a higher estimation of the logarithmic spectrum at the lowest cost and improve the validity of the spectral domain and the time domain Sex. The article tested several different W SS noise and different signal to noise ratio (SNR) situation, in the range of 3dB ~ 10dB, ANN method is significantly better than the use of speech signal energy and zero-crossing detection method, At the same time, the accuracy of other complex cepstrum matrix methods is also improved.