首页>
外文会议>Audio Engineering Society Convention
>Machine Learning Applied to Aspirated and Non-aspirated Allophone Classification - an Approach Based on Audio 'Fingerprinting'
【24h】
Machine Learning Applied to Aspirated and Non-aspirated Allophone Classification - an Approach Based on Audio 'Fingerprinting'
The purpose of this study is to involve both Convolutional Neural Networks and a typical learning algorithm in the allophone classification process. A list of words including aspirated and non-aspirated allophones pronounced by native and non-native English speakers is recorded and then edited and analyzed. Allophones extracted from English speakers' recordings are presented in the form of two-dimensional spectrogram images and used as input to train the Convolutional Neural Networks. Various settings of the spectral representation are analyzed to determine adequate option for the allophone classification. Then, testing is performed on the basis of non-native speakers' utterances. The same approach is repeated employing learning algorithm but based on feature vectors. The achieved classification results are promising as high accuracy is observed.
展开▼
机译:本研究的目的是涉及卷积神经网络和典型的学习算法在漫画中的分类过程中。记录包括本机和非原生扬声器发音的有吸气和非吸气的漫画的单词列表,然后编辑和分析。从英语扬声器录音中提取的alloChame以二维谱图图像的形式呈现,并用作培训卷积神经网络的输入。 Various settings of the spectral representation are analyzed to determine adequate option for the allophone classification.然后,在非母语扬声器的话语的基础上进行测试。重复采用学习算法的相同方法,但基于特征向量。所实现的分类结果是有前途的,因为观察到高精度。
展开▼