Loading...
Please wait, while we are loading the content...
Similar Documents
Feature Extraction Using Adaptive Restricted Boltzmann Machine for Dysarthric Speech Recognition
| Content Provider | Semantic Scholar |
|---|---|
| Author | Yuki, Takashima Toru, Nakashika Tetsuya, Takiguchi Yasuo, Ariki |
| Copyright Year | 2017 |
| Abstract | We investigate in this paper a feature extraction method for a person with an articulation disorder resulting from athetoid cerebral palsy. In the case of a person with this type of articulation disorder, the articulation of the utterances tends to become unstable due to the strain placed on the speech-related muscles. In our previous work, the feature extraction method using a convolutional neural network was proposed. In general, neural networks require sufficient amount of training data. However, generally speaking, the amount of speech data obtained from a person with an articulation disorder is limited because their burden is large due to strain on the speech muscles. Because the dysarthric speech fluctuates every utterance, it is difficult to obtain the correct alignment. In this paper, we propose a feature extraction method using adaptive restricted Boltzmann machine (ARBM). Because an ARBM is trained separating the speaker-independent parameters and the speaker-dependent parameters expressly, the amount of the training data can be reduced. The parameters of an ARBM are estimated using unsupervised learning. Therefore, it is not necesary to use incorrect label information. In this paper, we report our experimental results of speech recognition using the features extracted from our proposed method. |
| Starting Page | 321 |
| Ending Page | 326 |
| Page Count | 6 |
| File Format | PDF HTM / HTML |
| Volume Number | 116 |
| Alternate Webpage(s) | http://www.me.cs.scitec.kobe-u.ac.jp/~takigu/pdf/2017/EA2016-140.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
