排序方式: 共有3条查询结果,搜索用时 0 毫秒
1
1.
多语种情感语音的韵律特征分析和情感识别研究 总被引:3,自引:1,他引:2
韵律特征参数的变化是语音信号中情感信息主要体现。为了研究基于少量韵律特征的多语种语音样本情感识别的可行性,以提高情感识别系统对语种信息的鲁棒性,实验选取七种典型的情感状态,对指定句式下同一说话人在汉语、英语、日语多语种语音样本中的基频、能量、时间等韵律参数的动态特性进行统计分析。统计结果表明,不同语种情感语音样本的各种韵律特征参数的变化结构有较好的一致性。在这一结论基础上,利用主元素分析方法(PCA)对多语种混合样本进行了初步的情感识别实验,平均错误率为27.74%,最低识别错误率为11%。可见,通过基本的韵律参数可以实现对几种基本情感忽略语种信息的初步有效识别。 相似文献
2.
The conventional symmetry numbers σ_≠~' of activated complexes may lead to error in the rate constant expression of transition state theory, whereas the statistical factor ι~≠ or ι may violate the principle of detailed balance. A mathematically precise definition of the symmetry number σ_≠ of activated complex is given, i.e. σ_≠=_iN_4(?)/m, m is the number of physically distinct configurations of labelled transition state and N_i is the identical atoms in the activated complex. The identical atoms must belong to the same molecule of reactants and products. The present symmetry numbers σ_≠ of activated complexes assure not only obtaining correct rate constant expressions but also obeying the principle of detailed balance. It can be used with the statistical factor ι to construct the structures of transition states for unimolecular and bimolecular exchange reactions. 相似文献
3.
1