基于HMM与K-均值聚类的声目标识别

Approach Based on HMM and K-means Clustering to Identify Acoustic Target

提出一种隐马尔可夫模型和K-均值聚类混合模型的声目标识别方法。在该方法中,建立声信号的HMM,提取了声信号的线性预测系数(LPC)作为目标识别的参数,用K-均值算法对参数进行聚类,产生了训练和识别所用的特征向量。最后根据混合模型的识别算法判断声目标的类别。仿真结果表明:新的混合模型识别系统在声目标识别中具有可行性。

A recognition method based on HMM and K-means cluster is proposed through extracting LPC characteristic from acoustic target. The HMM was employed to compute the Viterbi output score and K-means algorithm is used as cluster LPC. Finally target is recognized by hybrid model, Simulation indicate this method is effective in target recognition.