首页>
外国专利>
A VOICE ACTIVITY DETECTING METHOD BASED ON A SUPPORT VECTOR MACHINESVM USING A POSTERIORI SNR, A PRIORI SNR AND A PREDICTED SNR AS A FEATURE VECTOR
A VOICE ACTIVITY DETECTING METHOD BASED ON A SUPPORT VECTOR MACHINESVM USING A POSTERIORI SNR, A PRIORI SNR AND A PREDICTED SNR AS A FEATURE VECTOR
展开▼
机译:基于后验SNR,先验SNR和预测SNR作为特征向量的基于支持向量机的语音活动检测方法
展开▼
页面导航
摘要
著录项
相似文献
摘要
A voice activity detecting method based on a support vector machine using a posteriori SNR(Signal-to-Noise Ratio), a priori SNR, and a predicted SNR as a feature vector is provided to improve voice detecting performance by combining the posteriori SNR, the priori SNR, and the predicted SNR with each other. A voice activity detecting method based on a support vector machine using a posteriori SNR, a priori SNR, and a predicted SNR as a feature vector includes: extracting a posteriori SNR, a priori SNR, and a predicted SNR from leaning voice data(310); combing the posteriori SNR, the priori SNR, and the predicted SNR with each other to produce a training feature vector(320); producing an SVM(Support Vector Machine) model obtaining an optimal weight vector and an optimal bias based on the produced training feature(330); extracting the posteriori SNR, the priori SNR, and the predicted SNR from voice data to be tested(340); combing the posteriori SNR, the priori SNR, and the predicted SNR with each other to produce a test feature vector(350); and applying the extracted test feature vector to the SVM model produced in a step of producing the SVM model to detect a voice(360).
展开▼