论文部分内容阅读
推导了基于独立分量分析(independent component analysis,ICA)的信号表征方法的有效性,并将基于ICA的信号特征分析应用到说话者识别任务中,通过计算机仿真实验证明该方法对非高斯语音信号受高斯噪声干扰时特征提取的有效性。最后的比较结果证明,基于ICA的信号特征,性能较传统的基于离散余弦变换(discrete cosinetransform,DCT)的Mel频率倒谱系数(mel-frequency cepstral coefficients,MFCC)特征有明显的提高。
The validity of signal characterization method based on independent component analysis (ICA) is deduced, and the signal characteristic analysis based on ICA is applied to the speaker recognition task. Computer simulation results show that this method is effective to non-Gaussian speech signal Effectiveness of Feature Extraction When Gaussian Noise Interferes. The final comparison results show that the characteristic of Mel-frequency cepstral coefficients (MFCC) based on discrete signal transform (discrete cosine transform) is obviously improved based on ICA signal characteristics.