http://tech.groups.yahoo.com/group/icsi-speech-tools/ icsi-speech-tools Fri, 14 Nov 2008 03:32:09 GMT qingqing.zhang http://tech.groups.yahoo.com/group/icsi-speech-tools/message/162 http://tech.groups.yahoo.com/group/icsi-speech-tools/message/162 Dear Adam, When I use qnstrn to get TRAP feature, it prints "(QN_InFtrStream_RandWindow.ftr1_file): WARNING - Segment 2292 has only 90 frames, which is too Fri, 14 Nov 2008 02:22:01 GMT qingqing.zhang http://tech.groups.yahoo.com/group/icsi-speech-tools/message/161 http://tech.groups.yahoo.com/group/icsi-speech-tools/message/161 Dear Adam, When I use qnstrn to get TRAP feature, it prints "(QN_InFtrStream_RandWindow.ftr1_file): WARNING - Segment 2292 has only 90 frames, which is too Thu, 13 Nov 2008 09:58:20 GMT qingqing.zhang http://tech.groups.yahoo.com/group/icsi-speech-tools/message/160 http://tech.groups.yahoo.com/group/icsi-speech-tools/message/160 Until now, PLP_MLP feature increased the performance accuracy of Mandarin CTS in our experiment. Now we want to try TRAPs into our system. If I use qnstrn to Mon, 03 Nov 2008 19:23:27 GMT Adam Janin http://tech.groups.yahoo.com/group/icsi-speech-tools/message/159 http://tech.groups.yahoo.com/group/icsi-speech-tools/message/159 ... I haven't run Mandarin CTS myself, so maybe somebody else on the list can answer in more detail, but for English, high 60s for frame ACCURACY (not error) Mon, 03 Nov 2008 19:21:15 GMT Adam Janin http://tech.groups.yahoo.com/group/icsi-speech-tools/message/158 http://tech.groups.yahoo.com/group/icsi-speech-tools/message/158 From your message, I assume you're using the MLP outputs as features to a GMM (so-called "tandem") rather than using the posteriors directly (so-called Mon, 03 Nov 2008 07:37:56 GMT qingqing.zhang http://tech.groups.yahoo.com/group/icsi-speech-tools/message/157 http://tech.groups.yahoo.com/group/icsi-speech-tools/message/157 There is another problem: When we use MLP feature on Mandarin CTS, the Frame Error Rate of training data was up to 60% after NN training. Does this a little Mon, 03 Nov 2008 03:26:03 GMT qingqing.zhang http://tech.groups.yahoo.com/group/icsi-speech-tools/message/156 http://tech.groups.yahoo.com/group/icsi-speech-tools/message/156 Dear Adam, I am so sorry that I did not explain the problem clearly. Actually, we found that MLP has the much larger dynamic range than PLP has (In the log Thu, 30 Oct 2008 16:58:41 GMT Adam Janin http://tech.groups.yahoo.com/group/icsi-speech-tools/message/155 http://tech.groups.yahoo.com/group/icsi-speech-tools/message/155 ... I'm not sure I understand. If you're using softmax as the output layer of the MLP, then all the values should be between 0.0 and 1.0. If you're not using Thu, 30 Oct 2008 07:26:01 GMT qingqing.zhang http://tech.groups.yahoo.com/group/icsi-speech-tools/message/154 http://tech.groups.yahoo.com/group/icsi-speech-tools/message/154 When I use MLP as my HMM features, the POSITIVE posteriori probabilities always occurred. The POSITIVE posteriori probabilities will break the decoding rule Tue, 30 Sep 2008 17:37:21 GMT Adam Janin http://tech.groups.yahoo.com/group/icsi-speech-tools/message/153 http://tech.groups.yahoo.com/group/icsi-speech-tools/message/153 ... Does your CV set start at segment 4802 by any chance? If so, that's what causes the message -- when the CV run starts, it seeks to the start of the CV run. Tue, 30 Sep 2008 17:23:19 GMT Adam Janin http://tech.groups.yahoo.com/group/icsi-speech-tools/message/152 http://tech.groups.yahoo.com/group/icsi-speech-tools/message/152 ... If you're computing normalization online and then try to seek to the middle of a feature stream, it has to back up to the start and read all the Tue, 30 Sep 2008 07:00:59 GMT qingqing.zhang http://tech.groups.yahoo.com/group/icsi-speech-tools/message/151 http://tech.groups.yahoo.com/group/icsi-speech-tools/message/151 When I run qnstrn with online CMS, it prints "./qnstrn(QN_InFtrStream_OnlNorm.ftr1_file): WARNING - slow seek to mid-seg (seg/fr 4802/81) with online norm" Wed, 24 Sep 2008 16:22:00 GMT Adam Janin http://tech.groups.yahoo.com/group/icsi-speech-tools/message/150 http://tech.groups.yahoo.com/group/icsi-speech-tools/message/150 ... I haven't looked at this stuff in some time, but dpwelib should have been included. You can also get it individually from: Wed, 24 Sep 2008 04:29:08 GMT qingqing.zhang http://tech.groups.yahoo.com/group/icsi-speech-tools/message/149 http://tech.groups.yahoo.com/group/icsi-speech-tools/message/149 When I executed ./labels2pfile, it printed "can't find package Dpwe_Utilfns while executing "package require Dpwe_Utilfns" (file "./labels2pfile" line 16)" Did Tue, 13 May 2008 23:56:06 GMT David Gelbart http://tech.groups.yahoo.com/group/icsi-speech-tools/message/148 http://tech.groups.yahoo.com/group/icsi-speech-tools/message/148 Hi, This week, we are planning to make noway version 2.9.3 available at ftp://ftp.icsi.berkeley.edu/pub/real/davidj/. The only change compared to version 2.9.2