研究了音频信息处理中一项重要的预处理工作:语音音乐分类。针对语音信号处理中遇到的实际问题,选择合适的音频特征和分类器来对音频数据进行语音和音乐分类。采用二级系统,选择优化低能量率(ModifiedLow Energy Ratio,MLER)以及梅尔频...研究了音频信息处理中一项重要的预处理工作:语音音乐分类。针对语音信号处理中遇到的实际问题,选择合适的音频特征和分类器来对音频数据进行语音和音乐分类。采用二级系统,选择优化低能量率(ModifiedLow Energy Ratio,MLER)以及梅尔频谱倒谱系数(Mel Frequency Cepstral Coefficients,MFCC)作为音频特征,通过贝叶斯分类和混合高斯分类器进行分类。最后,使用上下文分类器对分类结果进行修正。实验结果表明,这种分类方法准确率和速度都较好。展开更多
音频分类是音频信号处理中一项重要的预处理工作。该文描述了一种基于能量的分类方法,将音频信号分为语音和音乐2种类型。分类的过程分为3个阶段,首先计算优化低能量率MLER(modified low energy ratio)作为特征,然后利用初级分类器得到...音频分类是音频信号处理中一项重要的预处理工作。该文描述了一种基于能量的分类方法,将音频信号分为语音和音乐2种类型。分类的过程分为3个阶段,首先计算优化低能量率MLER(modified low energy ratio)作为特征,然后利用初级分类器得到初步分类的结果,最后利用音频类别的前后相关性,使用上下文分类器修正初始分类得到最终分类的结果。该文重点对MLER中参数的合理选取范围进行了讨论,并对传统的初始分类器作了改进,用非参数分类器和参数分类器代替原有的Bayes硬判决的方法,避免了由于门限选择不当所带来的分类错误。实验表明,使用参数分类器时,对纯语音和纯音乐分类效果很好,正确率达99%以上。展开更多
A novel VOF-type volume-tracking method for two-dimensional free-surface flows based on the unstructured triangular mesh is presented. Owing to the inherent merit of the unstructured triangular mesh in fitting curved ...A novel VOF-type volume-tracking method for two-dimensional free-surface flows based on the unstructured triangular mesh is presented. Owing to the inherent merit of the unstructured triangular mesh in fitting curved boundaries, this method can handle the free-surface problems with complex geometries accurately and directly, without introducing any complicated boundary treatment or artificial diffusion. The method solves the volume transport equation geometrically through the Modified Lagrangian-Eulerian Re-map (MLER) method, which is applied to advective fluid volumes. Moreover, the PLIC method is adopted to give a second-order reconstructed interface approximation. To validate this method, two advection tests were performed for the establishment of the accuracy and convergence rate of the solutions. Numerical results for these complex tests provide convincing evidence for the excellent solution quality and fidelity of the method.展开更多
To simulate two-dimensional free-surface flows with complex boundaries directly and accurately, a novel VOF (Volume-of-fluid) method based on unstructured quadrilateral mesh is presented. Without introducing any compl...To simulate two-dimensional free-surface flows with complex boundaries directly and accurately, a novel VOF (Volume-of-fluid) method based on unstructured quadrilateral mesh is presented. Without introducing any complicated boundary treatment or artificial diffusion, this method treated curved boundaries directly by utilizing the inherent merit of unstructured mesh in fitting curves. The PLIC (Piecewise Linear Interface Calculation) method was adopted to obtain a second-order accurate linearized reconstruction approximation and the MLER (Modified Lagrangian-Eulerian Re-map) method was introduced to advect fluid volumes on unstructured mesh. Moreover, an analytical relation for the interface’s line constant vs. the volume clipped by the interface was developed so as to improve the method’s efficiency. To validate this method, a comprehensive series of large straining advection tests were performed. Numerical results provide convincing evidences for the method’s high volume conservative accuracy and second-order shape error convergence rate. Also, a dramatic improvement on computational accuracy over its unstructured triangular mesh counterpart is checked.展开更多
文摘研究了音频信息处理中一项重要的预处理工作:语音音乐分类。针对语音信号处理中遇到的实际问题,选择合适的音频特征和分类器来对音频数据进行语音和音乐分类。采用二级系统,选择优化低能量率(ModifiedLow Energy Ratio,MLER)以及梅尔频谱倒谱系数(Mel Frequency Cepstral Coefficients,MFCC)作为音频特征,通过贝叶斯分类和混合高斯分类器进行分类。最后,使用上下文分类器对分类结果进行修正。实验结果表明,这种分类方法准确率和速度都较好。
文摘音频分类是音频信号处理中一项重要的预处理工作。该文描述了一种基于能量的分类方法,将音频信号分为语音和音乐2种类型。分类的过程分为3个阶段,首先计算优化低能量率MLER(modified low energy ratio)作为特征,然后利用初级分类器得到初步分类的结果,最后利用音频类别的前后相关性,使用上下文分类器修正初始分类得到最终分类的结果。该文重点对MLER中参数的合理选取范围进行了讨论,并对传统的初始分类器作了改进,用非参数分类器和参数分类器代替原有的Bayes硬判决的方法,避免了由于门限选择不当所带来的分类错误。实验表明,使用参数分类器时,对纯语音和纯音乐分类效果很好,正确率达99%以上。
文摘A novel VOF-type volume-tracking method for two-dimensional free-surface flows based on the unstructured triangular mesh is presented. Owing to the inherent merit of the unstructured triangular mesh in fitting curved boundaries, this method can handle the free-surface problems with complex geometries accurately and directly, without introducing any complicated boundary treatment or artificial diffusion. The method solves the volume transport equation geometrically through the Modified Lagrangian-Eulerian Re-map (MLER) method, which is applied to advective fluid volumes. Moreover, the PLIC method is adopted to give a second-order reconstructed interface approximation. To validate this method, two advection tests were performed for the establishment of the accuracy and convergence rate of the solutions. Numerical results for these complex tests provide convincing evidence for the excellent solution quality and fidelity of the method.
基金the National Natural Science Foundation ofChina under Grant No. 50779043, 50779045
文摘To simulate two-dimensional free-surface flows with complex boundaries directly and accurately, a novel VOF (Volume-of-fluid) method based on unstructured quadrilateral mesh is presented. Without introducing any complicated boundary treatment or artificial diffusion, this method treated curved boundaries directly by utilizing the inherent merit of unstructured mesh in fitting curves. The PLIC (Piecewise Linear Interface Calculation) method was adopted to obtain a second-order accurate linearized reconstruction approximation and the MLER (Modified Lagrangian-Eulerian Re-map) method was introduced to advect fluid volumes on unstructured mesh. Moreover, an analytical relation for the interface’s line constant vs. the volume clipped by the interface was developed so as to improve the method’s efficiency. To validate this method, a comprehensive series of large straining advection tests were performed. Numerical results provide convincing evidences for the method’s high volume conservative accuracy and second-order shape error convergence rate. Also, a dramatic improvement on computational accuracy over its unstructured triangular mesh counterpart is checked.