“130916-Zhiyong Zhang”版本间的差异
来自cslt Wiki
（以内容“Weekly Report: 1. Finished 23/30/40 fbanks DNN CMVN/noCMVN trainging and decoding. 2. Finished fbank_40 LDA:200/300/400 CMVN training and decoding. 2. Finished 100hs' ...”创建新页面） |
|||
| 第1行： | 第1行： | ||
Weekly Report: | Weekly Report: | ||
| + | |||
1. Finished 23/30/40 fbanks DNN CMVN/noCMVN trainging and decoding. | 1. Finished 23/30/40 fbanks DNN CMVN/noCMVN trainging and decoding. | ||
2. Finished fbank_40 LDA:200/300/400 CMVN training and decoding. | 2. Finished fbank_40 LDA:200/300/400 CMVN training and decoding. | ||
| 第5行： | 第6行： | ||
Target of next week: | Target of next week: | ||
| + | |||
1. To finish the 100hs' sparse DNN test based on 15db white noise dataset. | 1. To finish the 100hs' sparse DNN test based on 15db white noise dataset. | ||
2. To continue the training of 1000hs' fbank_40 LDA_200 DNN training. | 2. To continue the training of 1000hs' fbank_40 LDA_200 DNN training. | ||
3. To communicate with the professional patent writers to finish patent writing. | 3. To communicate with the professional patent writers to finish patent writing. | ||
2013年9月16日 (一) 07:07的版本
Weekly Report:
1. Finished 23/30/40 fbanks DNN CMVN/noCMVN trainging and decoding. 2. Finished fbank_40 LDA:200/300/400 CMVN training and decoding. 2. Finished 100hs' clean speech DNN model sparse DNN training and test.
Target of next week:
1. To finish the 100hs' sparse DNN test based on 15db white noise dataset. 2. To continue the training of 1000hs' fbank_40 LDA_200 DNN training. 3. To communicate with the professional patent writers to finish patent writing.
