基于加权特征补偿变换的说话人识别仿真研究

在噪声环境下,为提高说话人识别系统的鲁棒性,需要对系统进行各种抗噪声处理。从特征值处理和模型补偿两方面考虑,提出了基于高斯混合模型的加权特征补偿变换的抗噪声方法。一方面根据帧信噪比对特征值的贡献进行加权;另一方面根据说话人识别的声学特性对模型输出的似然得分进行补偿变换。实验结果表明,不管在平稳噪声还是非平稳噪声环境下,该方法都能取得较好的识别效果,并进一步增强系统的鲁棒性。这也说明加权特征补偿变换是降低噪声和干扰的影响、提高系统识别率和鲁棒性的有效方法。

Speaker Recognition Based on Weighted Features Compensation Transformation and Its Simulation Study

Diversified methods of decreasing the influence of noise have appeared to improve the performance of speaker recognition system in noise. Based on feature processing and model compensation, a weighted features compensation transformation method based on GMM for robust speaker verification was proposed. In the method, the scores of features were weighted through frame SNR, while the frame likelihood probabilities were transformed based on the acoustic characteristic of speaker recognition system. The experiment shows that the system could acquire better results and the robustness could be improved with the new method. The result shows weighted normalization compensation transformation should be adopted for canceling the influence of variations in noise, model mismatch and improving the recognition rate and robustness.