摘要
针对网球视频用于网球教学的问题,文中提出了一种基于隐马尔科夫(HMM)的网球音频数据语义分析方法。该方法首先对音频数据进行分帧处理,并提取短时过零率、短时平均能量、梅尔频率倒谱系数和差分倒谱系数构成特征向量,然后基于HMM构建音频数据语义分析模型,并利用Baum-Welch算法完成模型计算,以状态概率最大对应的类别作为语义分析结果。实验结果表明,该方法能够有效实现网球音频数据的语义特征,且具有较高的识别准确率。
Aiming at the problem in tennis teaching by tennis video,a semantic analysis method based on hidden Markov model(HMM)for tennis audio data is proposed.Firstly,the audio data is frameed to extract the short-time zero-crossing rate,the short-term average energy,the Mel frequency cepstrum coefficient and the differential cepstral coefficient to form the feature vector.Then a semantic analysis model based on HMM is constructed.The Baum-Welch algorithm is used to complete the model calculation,and the category with the highest state probability is used as the semantic analysis result.The experimental results show that the proposed method can effectively realize the semantic features of tennis audio data and has higher recognition accuracy.
作者
徐翠萍
XU Cui-ping(College of Mechanical Engineering,Shaanxi Polytechnic institute,Xianyang 712000,Shaanxi Province,China)
出处
《信息技术》
2019年第8期103-106,111,共5页
Information Technology