基于特征分析的环境声音事件识别算法

摘要

The algorithm on Acoustic Event Classification(AEC) always uses traditional speech classification model at present, but to environment acoustic event, this algorithm has lower correct rate and bad stability. This paper puts forward an environment acoustic event classification algorithm based on feature analysis. It makes a definition to environment acoustic event. It analyzes familiar features of sound. It makes a classification to four typical kinds environment acoustic event just using features. Experiment proves that this method is better than traditional algorithm in correct rate and stability.%对于环境声音事件,传统语音识别算法的识别效率低、稳定性差.为此,提出一种基于特征分析的环境声音事件识别算法.定义环境声音事件,分析常用的声音特征,不使用分类模型,仅利用特征对4种典型的环境声音事件进行分类.实验证明,该算法在识别率和稳定性上都优于传统识别算法,能够完成分类任务.