Voice Recognition Implementation
-
@Naveen_D
It seems like a good tut.When you compile you will get the lib
and this lib you will use in the real project. -
@mrjj In qt creator i need to compile?
one more question, since i am new i want know what this sudo make install will do ?
-
Im not sure if it comes with a .pro file?
Its should be clear from build instructions what to do
on each platform.sudo make install
That will copy the "result" (exe, .a ) to a place so its installed.The tut you found is not Qt related so its with normal linux toolchain
for CMU-Sphinx.He starts with
"When I installed SPHINX for the first time in September 2015, it was not a fun experience."So be prepared to really read what he does and read docs for it. its not trivial to build.
-
The tut you found is not Qt related so its with normal linux toolchain
for CMU-Sphinx.He starts with
"When I installed SPHINX for the first time in September 2015, it was not a fun experience."So be prepared to really read what he does and read docs for it. its not trivial to build.
-
First you get it to build then
use in Qt. its in 2 steps.
You really MUST read docs slow and carefully ,
else u miss a step and it wont work.So first step is to get it to build using the tuts.
-
First you get it to build then
use in Qt. its in 2 steps.
You really MUST read docs slow and carefully ,
else u miss a step and it wont work.So first step is to get it to build using the tuts.
-
First you get it to build then
use in Qt. its in 2 steps.
You really MUST read docs slow and carefully ,
else u miss a step and it wont work.So first step is to get it to build using the tuts.
-
The tut you found seemed good?
http://jrmeyer.github.io/installation/2016/01/09/Installing-CMU-Sphinx-on-Ubuntu.htmlSo if you have build the so files you are to use them
This part mostly fails due to paths so make sure u check.You can open .pro file then, in open file, right click and select
Add library. Fill it out and make sure!! its correct.
Linux lib, shared etc.http://doc.qt.io/qtcreator/creator-project-qmake-libraries.html
Then you should be able to link against it.
-
https://github.com/cmusphinx/pocketsphinx
Same story with pocket
http://doc.qt.io/qtcreator/creator-project-qmake-libraries.html
also MAKE 100% sure you did as it says
"Prerequisites
You must have SphinxBase, which you can download from http://cmusphinx.sourceforge.net. Download and unpack it to the same parent directory as PocketSphinx, so that the configure script and project files can find it. On Windows, you will need to rename 'sphinxbase-X.Y' (where X.Y is the SphinxBase version number) to simply 'sphinxbase' for this to work."
-
https://github.com/cmusphinx/pocketsphinx
Same story with pocket
http://doc.qt.io/qtcreator/creator-project-qmake-libraries.html
also MAKE 100% sure you did as it says
"Prerequisites
You must have SphinxBase, which you can download from http://cmusphinx.sourceforge.net. Download and unpack it to the same parent directory as PocketSphinx, so that the configure script and project files can find it. On Windows, you will need to rename 'sphinxbase-X.Y' (where X.Y is the SphinxBase version number) to simply 'sphinxbase' for this to work."
@mrjj i want to attach the screen shot of what files i have got after running the steps in that link http://jrmeyer.github.io/installation/2016/01/09/Installing-CMU-Sphinx-on-Ubuntu.html for both sphinxbase and pocketsphinx but i qm not able to do that...how can we attach the screen shots here ???
-
Use external site like postimage.org and paste link here or use
 -
Use external site like postimage.org and paste link here or use
 -
Well if all works and says no "error" of any kind,
I would try one of the existing examples
and see if it works.Then I would start think about how to use it in Qt.
@mrjj after running the command pocketsphinx_continuous i get the following result
ubuntu@ub:~/Desktop/sphinx-source/pocketsphinx$ pocketsphinx_continuous
ERROR: "cmd_ln.c", line 682: No arguments given, available options are:
Arguments list definition:
[NAME] [DEFLT] [DESCR]
-adcdev Name of audio device to use for input.
-agc none Automatic gain control for c0 ('max', 'emax', 'noise', or 'none')
-agcthresh 2.0 Initial threshold for automatic gain control
-allphone Perform phoneme decoding with phonetic lm
-allphone_ci no Perform phoneme decoding with phonetic lm and context-independent units only
-alpha 0.97 Preemphasis parameter
-argfile Argument file giving extra arguments.
-ascale 20.0 Inverse of acoustic model scale for confidence score calculation
-aw 1 Inverse weight applied to acoustic scores.
-backtrace no Print results and backtraces to log.
-beam 1e-48 Beam width applied to every frame in Viterbi search (smaller values mean wider beam)
-bestpath yes Run bestpath (Dijkstra) search over word lattice (3rd pass)
-bestpathlw 9.5 Language model probability weight for bestpath search
-ceplen 13 Number of components in the input feature vector
-cmn live Cepstral mean normalization scheme ('live', 'batch', or 'none')
-cmninit 40,3,-1 Initial values (comma-separated) for cepstral mean when 'live' is used
-compallsen no Compute all senone scores in every frame (can be faster when there are many senones)
-debug Verbosity level for debugging messages
-dict Main pronunciation dictionary (lexicon) input file
-dictcase no Dictionary is case sensitive (NOTE: case insensitivity applies to ASCII characters only)
-dither no Add 1/2-bit noise
-doublebw no Use double bandwidth filters (same center freq)
-ds 1 Frame GMM computation downsampling ratio
-fdict Noise word pronunciation dictionary input file
-feat 1s_c_d_dd Feature stream type, depends on the acoustic model
-featparams File containing feature extraction parameters.
-fillprob 1e-8 Filler word transition probability
-frate 100 Frame rate
-fsg Sphinx format finite state grammar file
-fsgusealtpron yes Add alternate pronunciations to FSG
-fsgusefiller yes Insert filler words at each state.
-fwdflat yes Run forward flat-lexicon search over word lattice (2nd pass)
-fwdflatbeam 1e-64 Beam width applied to every frame in second-pass flat search
-fwdflatefwid 4 Minimum number of end frames for a word to be searched in fwdflat search
-fwdflatlw 8.5 Language model probability weight for flat lexicon (2nd pass) decoding
-fwdflatsfwin 25 Window of frames in lattice to search for successor words in fwdflat search
-fwdflatwbeam 7e-29 Beam width applied to word exits in second-pass flat search
-fwdtree yes Run forward lexicon-tree search (1st pass)
-hmm Directory containing acoustic model files.
-infile Audio file to transcribe.
-inmic no Transcribe audio from microphone.
-input_endian little Endianness of input data, big or little, ignored if NIST or MS Wav
-jsgf JSGF grammar file
-keyphrase Keyphrase to spot
-kws A file with keyphrases to spot, one per line
-kws_delay 10 Delay to wait for best detection score
-kws_plp 1e-1 Phone loop probability for keyphrase spotting
-kws_threshold 1 Threshold for p(hyp)/p(alternatives) ratio
-latsize 5000 Initial backpointer table size
-lda File containing transformation matrix to be applied to features (single-stream features only)
-ldadim 0 Dimensionality of output of feature transformation (0 to use entire matrix)
-lifter 0 Length of sin-curve for liftering, or 0 for no liftering.
-lm Word trigram language model input file
-lmctl Specify a set of language model
-lmname Which language model in -lmctl to use by default
-logbase 1.0001 Base in which all log-likelihoods calculated
-logfn File to write log messages in
-logspec no Write out logspectral files instead of cepstra
-lowerf 133.33334 Lower edge of filters
-lpbeam 1e-40 Beam width applied to last phone in words
-lponlybeam 7e-29 Beam width applied to last phone in single-phone words
-lw 6.5 Language model probability weight
-maxhmmpf 30000 Maximum number of active HMMs to maintain at each frame (or -1 for no pruning)
-maxwpf -1 Maximum number of distinct word exits at each frame (or -1 for no pruning)
-mdef Model definition input file
-mean Mixture gaussian means input file
-mfclogdir Directory to log feature files to
-min_endfr 0 Nodes ignored in lattice construction if they persist for fewer than N frames
-mixw Senone mixture weights input file (uncompressed)
-mixwfloor 0.0000001 Senone mixture weights floor (applied to data from -mixw file)
-mllr MLLR transformation to apply to means and variances
-mmap yes Use memory-mapped I/O (if possible) for model files
-ncep 13 Number of cep coefficients
-nfft 512 Size of FFT
-nfilt 40 Number of filter banks
-nwpen 1.0 New word transition penalty
-pbeam 1e-48 Beam width applied to phone transitions
-pip 1.0 Phone insertion penalty
-pl_beam 1e-10 Beam width applied to phone loop search for lookahead
-pl_pbeam 1e-10 Beam width applied to phone loop transitions for lookahead
-pl_pip 1.0 Phone insertion penalty for phone loop
-pl_weight 3.0 Weight for phoneme lookahead penalties
-pl_window 5 Phoneme lookahead window size, in frames
-rawlogdir Directory to log raw audio files to
-remove_dc no Remove DC offset from each frame
-remove_noise yes Remove noise with spectral subtraction in mel-energies
-remove_silence yes Enables VAD, removes silence frames from processing
-round_filters yes Round mel filter frequencies to DFT points
-samprate 16000 Sampling rate
-seed -1 Seed for random number generator; if less than zero, pick our own
-sendump Senone dump (compressed mixture weights) input file
-senlogdir Directory to log senone score files to
-senmgau Senone to codebook mapping input file (usually not needed)
-silprob 0.005 Silence word transition probability
-smoothspec no Write out cepstral-smoothed logspectral files
-svspec Subvector specification (e.g., 24,0-11/25,12-23/26-38 or 0-12/13-25/26-38)
-time no Print word times in file transcription.
-tmat HMM state transition matrix input file
-tmatfloor 0.0001 HMM state transition probability floor (applied to -tmat file)
-topn 4 Maximum number of top Gaussians to use in scoring.
-topn_beam 0 Beam width used to determine top-N Gaussians (or a list, per-feature)
-toprule Start rule for JSGF (first public rule is default)
-transform legacy Which type of transform to use to calculate cepstra (legacy, dct, or htk)
-unit_area yes Normalize mel filters to unit area
-upperf 6855.4976 Upper edge of filters
-uw 1.0 Unigram weight
-vad_postspeech 50 Num of silence frames to keep after from speech to silence.
-vad_prespeech 20 Num of speech frames to keep before silence to speech.
-vad_startspeech 10 Num of speech frames to trigger vad from silence to speech.
-vad_threshold 2.0 Threshold for decision between noise and silence frames. Log-ratio between signal level and noise level.
-var Mixture gaussian variances input file
-varfloor 0.0001 Mixture gaussian variance floor (applied to data from -var file)
-varnorm no Variance normalize each utterance (only if CMN == current)
-verbose no Show input filenames
-warp_params Parameters defining the warping function
-warp_type inverse_linear Warping function type (or shape)
-wbeam 7e-29 Beam width applied to word exits
-wip 0.65 Word insertion penalty
-wlen 0.025625 Hamming window lengthINFO: continuous.c(295): Specify '-infile <file.wav>' to recognize from file or '-inmic yes' to recognize from microphone.
ubuntu@ub:~/Desktop/sphinx-source/pocketsphinx$ -
Does look like the tuts so I think its working :)
\o/ good work -
@mrjj after running the command pocketsphinx_continuous i get the following result
ubuntu@ub:~/Desktop/sphinx-source/pocketsphinx$ pocketsphinx_continuous
ERROR: "cmd_ln.c", line 682: No arguments given, available options are:
Arguments list definition:
[NAME] [DEFLT] [DESCR]
-adcdev Name of audio device to use for input.
-agc none Automatic gain control for c0 ('max', 'emax', 'noise', or 'none')
-agcthresh 2.0 Initial threshold for automatic gain control
-allphone Perform phoneme decoding with phonetic lm
-allphone_ci no Perform phoneme decoding with phonetic lm and context-independent units only
-alpha 0.97 Preemphasis parameter
-argfile Argument file giving extra arguments.
-ascale 20.0 Inverse of acoustic model scale for confidence score calculation
-aw 1 Inverse weight applied to acoustic scores.
-backtrace no Print results and backtraces to log.
-beam 1e-48 Beam width applied to every frame in Viterbi search (smaller values mean wider beam)
-bestpath yes Run bestpath (Dijkstra) search over word lattice (3rd pass)
-bestpathlw 9.5 Language model probability weight for bestpath search
-ceplen 13 Number of components in the input feature vector
-cmn live Cepstral mean normalization scheme ('live', 'batch', or 'none')
-cmninit 40,3,-1 Initial values (comma-separated) for cepstral mean when 'live' is used
-compallsen no Compute all senone scores in every frame (can be faster when there are many senones)
-debug Verbosity level for debugging messages
-dict Main pronunciation dictionary (lexicon) input file
-dictcase no Dictionary is case sensitive (NOTE: case insensitivity applies to ASCII characters only)
-dither no Add 1/2-bit noise
-doublebw no Use double bandwidth filters (same center freq)
-ds 1 Frame GMM computation downsampling ratio
-fdict Noise word pronunciation dictionary input file
-feat 1s_c_d_dd Feature stream type, depends on the acoustic model
-featparams File containing feature extraction parameters.
-fillprob 1e-8 Filler word transition probability
-frate 100 Frame rate
-fsg Sphinx format finite state grammar file
-fsgusealtpron yes Add alternate pronunciations to FSG
-fsgusefiller yes Insert filler words at each state.
-fwdflat yes Run forward flat-lexicon search over word lattice (2nd pass)
-fwdflatbeam 1e-64 Beam width applied to every frame in second-pass flat search
-fwdflatefwid 4 Minimum number of end frames for a word to be searched in fwdflat search
-fwdflatlw 8.5 Language model probability weight for flat lexicon (2nd pass) decoding
-fwdflatsfwin 25 Window of frames in lattice to search for successor words in fwdflat search
-fwdflatwbeam 7e-29 Beam width applied to word exits in second-pass flat search
-fwdtree yes Run forward lexicon-tree search (1st pass)
-hmm Directory containing acoustic model files.
-infile Audio file to transcribe.
-inmic no Transcribe audio from microphone.
-input_endian little Endianness of input data, big or little, ignored if NIST or MS Wav
-jsgf JSGF grammar file
-keyphrase Keyphrase to spot
-kws A file with keyphrases to spot, one per line
-kws_delay 10 Delay to wait for best detection score
-kws_plp 1e-1 Phone loop probability for keyphrase spotting
-kws_threshold 1 Threshold for p(hyp)/p(alternatives) ratio
-latsize 5000 Initial backpointer table size
-lda File containing transformation matrix to be applied to features (single-stream features only)
-ldadim 0 Dimensionality of output of feature transformation (0 to use entire matrix)
-lifter 0 Length of sin-curve for liftering, or 0 for no liftering.
-lm Word trigram language model input file
-lmctl Specify a set of language model
-lmname Which language model in -lmctl to use by default
-logbase 1.0001 Base in which all log-likelihoods calculated
-logfn File to write log messages in
-logspec no Write out logspectral files instead of cepstra
-lowerf 133.33334 Lower edge of filters
-lpbeam 1e-40 Beam width applied to last phone in words
-lponlybeam 7e-29 Beam width applied to last phone in single-phone words
-lw 6.5 Language model probability weight
-maxhmmpf 30000 Maximum number of active HMMs to maintain at each frame (or -1 for no pruning)
-maxwpf -1 Maximum number of distinct word exits at each frame (or -1 for no pruning)
-mdef Model definition input file
-mean Mixture gaussian means input file
-mfclogdir Directory to log feature files to
-min_endfr 0 Nodes ignored in lattice construction if they persist for fewer than N frames
-mixw Senone mixture weights input file (uncompressed)
-mixwfloor 0.0000001 Senone mixture weights floor (applied to data from -mixw file)
-mllr MLLR transformation to apply to means and variances
-mmap yes Use memory-mapped I/O (if possible) for model files
-ncep 13 Number of cep coefficients
-nfft 512 Size of FFT
-nfilt 40 Number of filter banks
-nwpen 1.0 New word transition penalty
-pbeam 1e-48 Beam width applied to phone transitions
-pip 1.0 Phone insertion penalty
-pl_beam 1e-10 Beam width applied to phone loop search for lookahead
-pl_pbeam 1e-10 Beam width applied to phone loop transitions for lookahead
-pl_pip 1.0 Phone insertion penalty for phone loop
-pl_weight 3.0 Weight for phoneme lookahead penalties
-pl_window 5 Phoneme lookahead window size, in frames
-rawlogdir Directory to log raw audio files to
-remove_dc no Remove DC offset from each frame
-remove_noise yes Remove noise with spectral subtraction in mel-energies
-remove_silence yes Enables VAD, removes silence frames from processing
-round_filters yes Round mel filter frequencies to DFT points
-samprate 16000 Sampling rate
-seed -1 Seed for random number generator; if less than zero, pick our own
-sendump Senone dump (compressed mixture weights) input file
-senlogdir Directory to log senone score files to
-senmgau Senone to codebook mapping input file (usually not needed)
-silprob 0.005 Silence word transition probability
-smoothspec no Write out cepstral-smoothed logspectral files
-svspec Subvector specification (e.g., 24,0-11/25,12-23/26-38 or 0-12/13-25/26-38)
-time no Print word times in file transcription.
-tmat HMM state transition matrix input file
-tmatfloor 0.0001 HMM state transition probability floor (applied to -tmat file)
-topn 4 Maximum number of top Gaussians to use in scoring.
-topn_beam 0 Beam width used to determine top-N Gaussians (or a list, per-feature)
-toprule Start rule for JSGF (first public rule is default)
-transform legacy Which type of transform to use to calculate cepstra (legacy, dct, or htk)
-unit_area yes Normalize mel filters to unit area
-upperf 6855.4976 Upper edge of filters
-uw 1.0 Unigram weight
-vad_postspeech 50 Num of silence frames to keep after from speech to silence.
-vad_prespeech 20 Num of speech frames to keep before silence to speech.
-vad_startspeech 10 Num of speech frames to trigger vad from silence to speech.
-vad_threshold 2.0 Threshold for decision between noise and silence frames. Log-ratio between signal level and noise level.
-var Mixture gaussian variances input file
-varfloor 0.0001 Mixture gaussian variance floor (applied to data from -var file)
-varnorm no Variance normalize each utterance (only if CMN == current)
-verbose no Show input filenames
-warp_params Parameters defining the warping function
-warp_type inverse_linear Warping function type (or shape)
-wbeam 7e-29 Beam width applied to word exits
-wip 0.65 Word insertion penalty
-wlen 0.025625 Hamming window lengthINFO: continuous.c(295): Specify '-infile <file.wav>' to recognize from file or '-inmic yes' to recognize from microphone.
ubuntu@ub:~/Desktop/sphinx-source/pocketsphinx$ -
@mrjj u said I would try one of the existing examples
and see if it works. where i can get existing examples ? -
@Naveen_D said in Voice Recognition Implementation:
pocketsphinx
http://cmusphinx.sourceforge.net/wiki/tutorialpocketsphinx
There is a Basic Usage (hello world) sample.
That should do it :) -
@Naveen_D said in Voice Recognition Implementation:
pocketsphinx
http://cmusphinx.sourceforge.net/wiki/tutorialpocketsphinx
There is a Basic Usage (hello world) sample.
That should do it :) -
@mrjj hi,
When run this command
pocketsphinx_continuous
-hmm /usr/share/pocketsphinx/model/hmm/en_US/hub4wsj_sc_8k
-dict /usr/share/pocketsphinx/model/lm/en_US/cmu07a.dic
-lm /usr/share/pocketsphinx/model/lm/en_US/hub4.5000.DMPi get this output can u pls tell me wats the prb here...
INFO: continuous.c(295): Specify '-infile <file.wav>' to recognize from file or '-inmic yes' to recognize from microphone.
