摘要
对基于相关向量机和矢量量化的语音识别算法模型进行了一系列的研究。与支持向量机识别算法相比,该算法基于贝叶斯统计模型理论,能够给出样本属于某一类的后验概率;而且,该算法充分利用了相关向量机所具有的高泛化性、核函数功能和结果的高稀疏性。基于矢量量化的特征提取仿真表明,该算法在减少相对误差和计算量方面有较大的优势。
A series of studies on speech recognition algorithm based on relevance vector machine(RVM)and vector quantization(VQ)are proposed.The sparseness and prediction probability of RVM make the algorithm suitable for speech recognition in applications.In contrast to the support vector machine(SVM)algorithm,the approach is based on the theory of Bayesian statistical model,giving the posterior probability of samples belonging to one kind.Moreover,the algorithm combines the high generalization,kernel tricks,and sparser performance of RVM to generate more robust classification results and to reduce the computational complexity.The simulations on feature extraction using VQ show that the proposed algorithm outperforms the other systems on reducing the relative error rates and reducing the computational complexity in high dimensionality space and big scale data.
出处
《黑龙江大学工程学报》
2014年第4期52-56,63,共6页
Journal of Engineering of Heilongjiang University
基金
中国船舶重工集团公司709所科技合作项目(J081113012)
哈尔滨市优秀学科带头人基金项目(RC2013XK009003)
关键词
语音识别
相关向量机
矢量量化
贝叶斯概率模型
支持向量机
speech recognition
relevance vector machine
vector quantization
Bayesian statistical model
support vector machine