Investigation on acoustic modeling with different phoneme set for continuous Lhasa Tibetan recognition based on DNN method.
Hongcui WangKuntharrgyal KhyuruJian LiGuanyu LiJianwu DangLixia HuangPublished in: APSIPA (2016)
Keyphrases
- method finds
- preprocessing
- cost function
- input data
- recognition algorithm
- pattern recognition
- high accuracy
- feature set
- computational cost
- neural network
- support vector machine
- acoustic signal
- high recognition accuracy
- recognition process
- segmentation method
- probabilistic model
- significant improvement
- pairwise
- objective function