Simplerec tutorial
From Sphinx
- Complete Sphinx3 Speech Recognition Quick Start because speech data files will be needed for this tutorial.
- Install sphinx3 trunk (version 0.7). NOTE: sphinx3 v. 0.6.3 will not work for this tutorial.
- Complete PortAudioPlayRec tutorial
- Download SphinxSimpleRec.zip to /usr/src and extract.
- Edit makefile
- Near the top of the file, add the line
- CMU_ROOT=$SPHINX_ROOT/sphinx3
- add -I/usr/src/sphinx/sphinxbase/include to INC variable
- Near the top of the file, add the line
- Edit wsj_5k_rec.cfg
- delete -mdef, -mean, -var, -mixw, -tmat variables and their values
- add -hmm $SPHINX_ROOT/sphinx3/model/hmm/hub4_cd_continuous_8gau_1s_c_d_dd
- replace the -dict, -fdict, -lm values with the values used in Sphinx3 Speech Recognition Quick Start
- delete -feat, -lw, -wip, -beam, -wbeam, -pbeam, -maxhmmpf, -maxwpf, -maxhistpf, and -outlatdir params
- Build
make
- If make fails, you may have to edit /SphinxSimpleRec.cpp, line 66 thus:
if (cmd_ln_parse_file(S3_DECODE_ARG_DEFS, argv[1] , 0) )
- Run the software
./SphinxSimpleRec wsj_5k_rec.cfg
Hit enter, wait for the beep to finish, then speak one of the words in the vocubulary: hello, please, or yellow.
