采用潜在概率语义模型和K近邻分类器的音频分类算法

Audio Classification Algorithm Using Probabilistic Latent Semantic Models and K Nearest Neighbor Classifier

提出一种基于潜在概率语义(PLSA)模型和K近邻分类器的音频分类算法.首先,将信号特征向量送入潜在概率语义模型中训练,获得声音主题词袋模型;然后,使用K近邻分类器(KNN)进行分类.实验结果表明:与传统的K近邻分类算法相比,提出的算法在分类效果上有较明显的改善.

The paper proposed an audio classification algorithm based on probabilistic latent semantic analysis model（PLSA）and K-nearest neighbor classifiers（KNN）.The algorithm first feed the audio signal feature vector into the PLSA model training to get a bag of sound frames models,then classify with the KNN classifier.Experimental results showed that the proposed classification algorithm has better classification effect compared with the traditional KNN algorithm.