Le Gia Anh Quy

Hi, I understand: 1. Linguistic questions are generated manually or automatically to cluster phones into groups. 2. A binary decision tree is built for each state of a phone (CI phone). My questions: 1. How do we can build decision trees for a CI phone when it does not have context dependency, e.g: left or right context? 2. Given a 3 state CI phone with 3 decision trees. Do those trees have the same set of linguistic questions for that specific phone? If yes, what is the point to have 3 similar decision...

Hi, I understand: 1. Linguistic questions are generated manually or automatically to cluster phones into groups. 2. A binary decision tree is built for each state of a phone (CI phone). My questions: 1. How do we can build decision trees for a CI phone if it does not have context dependency, e.g: left or right context? 2. Given a 3 state CI phone with 3 decision trees. Do those trees have the same set of linguistic questions for that specific phone? If yes, what is the point to have 3 similar decision...

Hello, I'm currently finding documentations about feature extraction algorithm used in the front-end of CMU Sphinx Architecture. What is the latest algorithm? I've recently found that PNCC is used. Is it true?

Current configuration: [NAME] [DEFLT] [VALUE] -alpha 0.97 9.700000e-01 -argfile vi/feat.params -blocksize 2048 2048 -build_outdirs yes yes -c vivosadp.fileids -cep2spec no no -di . -dither no no -do . -doublebw no no -ei wav -eo mfc -example no no -frate 100 100 -help no no -i -input_endian little little -lifter 0 0 -logspec no no -lowerf 133.33334 1.333333e+02 -mach_endian little little -mswav no yes -ncep 13 13 -nchans 1 1 -nfft 512 512 -nfilt 40 40 -nist no no -npart 0 0 -nskip 0 0 -o -ofmt sphinx...

I found that I used too much phoneme while it must be lower than 256

Hi, I have this error while training acoustic model: FATAL: "senone.c", line 213: fopen(/home/quylg/anaconda3/envs/Quy/asr/fpt/model_parameters/fpt.cd_ptm_4000_delinterp/mixture_weights,rb) failed Then, I check the deletedintrep-4000.log, I found: delint: main.c:575: rd_interp_wr: Assertion `accum_dir[i] != NULL' failed. My training data is about 17 hours, 5444 vocabularies, 253 phones. This error occurs because I used too much senone, doesn't it?

I'm also get this ERROR. Anyone solve this?

Well, I've found the solution myself. For anyone have the same issue, I rebuild my language model using CMUCLTK instead of using web-service LMTool