Sponsor:

Your company here, and a link to your site. Click to find out more.

pocketsphinx - Man Page

Run speech recognition on audio data

Synopsis

pocketsphinx [ options... ] [ live | single | help | soxflags ] INPUTS...

Description

The ‘pocketsphinx’ command-line program reads single-channel 16-bit PCM audio one or more input files (or ‘-’ to read from standard input), and attemps to recognize speech in it using the default acoustic and language model. The input files can be raw audio, WAV, or NIST Sphere files, though some of these may not be recognized properly.  It accepts a large number of options which you probably don't care about, and a command which defaults to ‘live’. The commands are as follows:

help

Print a long list of those options you don't care about.

config

Dump configuration as JSON to standard output (can be loaded with the ‘-config’ option).

live

Detect speech segments in input files, run recognition on them (using those options you don't care about), and write the results to standard output in line-delimited JSON. I realize this isn't the prettiest format, but it sure beats XML. Each line contains a JSON object with these fields, which have short names to make the lines more readable:

"b": Start time in seconds, from the beginning of the stream

"d": Duration in seconds

"p": Estimated probability of the recognition result, i.e. a number between 0 and 1 which may be used as a confidence score

"t": Full text of recognition result

"w": List of segments (usually words), each of which in turn contains the ‘b’, ‘d’, ‘p’, and ‘t’ fields, for start, end, probability, and the text of the word. In the future we may also support hierarchical results in which case ‘w’ could be present.

single

Recognize the input as a single utterance, and write a JSON object in the same format described above.

align

Align a single input file (or ‘-’ for standard input) to a word sequence, and write a JSON object in the same format described above. The first positional argument is the input, and all subsequent ones are concatenated to make the text, to avoid surprises if you forget to quote it.  You are responsible for normalizing the text to remove punctuation, uppercase, centipedes, etc. For example:

    pocketsphinx align goforward.wav "go forward ten meters"

By default, only word-level alignment is done.  To get phone alignments, pass `-phone_align yes` in the flags, e.g.:

    pocketsphinx -phone_align yes align audio.wav $text

This will make not particularly readable output, but you can use jq (https://stedolan.github.io/jq/) to clean it up.  For example, you can get just the word names and start times like this:

    pocketsphinx align audio.wav $text | jq '.w[]|[.t,.b]'

Or you could get the phone names and durations like this:

    pocketsphinx -phone_align yes align audio.wav $text | jq '.w[]|.w[]|[.t,.d]'

There are many, many other possibilities, of course.

help

Print a usage and help text with a list of possible arguments.

soxflags

Return arguments to ‘sox’ which will create the appropriate input format. Note that because the ‘sox’ command-line is slightly quirky these must always come after the filename or ‘-d’ (which tells ‘sox’ to read from the microphone). You can run live recognition like this:

    sox -d $(pocketsphinx soxflags) | pocketsphinx -

or decode from a file named "audio.mp3" like this:

    sox audio.mp3 $(pocketsphinx soxflags) | pocketsphinx -

By default only errors are printed to standard error, but if you want more information you can pass ‘-loglevel INFO’. Partial results are not printed, maybe they will be in the future, but don't hold your breath. Force-alignment is likely to be supported soon, however.

Options

-agc

Automatic gain control for c0 ('max', 'emax', 'noise', or 'none')

-agcthresh

Initial threshold for automatic gain control

-allphone

phoneme decoding with phonetic lm (given here)

-allphone_ci

Perform phoneme decoding with phonetic lm and context-independent units only

-alpha

Preemphasis parameter

-ascale

Inverse of acoustic model scale for confidence score calculation

-aw

Inverse weight applied to acoustic scores.

-backtrace

Print results and backtraces to log.

-beam

Beam width applied to every frame in Viterbi search (smaller values mean wider beam)

-bestpath

Run bestpath (Dijkstra) search over word lattice (3rd pass)

-bestpathlw

Language model probability weight for bestpath search

-ceplen

Number of components in the input feature vector

-cmn

Cepstral mean normalization scheme ('live', 'batch', or 'none')

-cmninit

Initial values (comma-separated) for cepstral mean when 'live' is used

-compallsen

Compute all senone scores in every frame (can be faster when there are many senones)

-dict

pronunciation dictionary (lexicon) input file

-dictcase

Dictionary is case sensitive (NOTE: case insensitivity applies to ASCII characters only)

-dither

Add 1/2-bit noise

-doublebw

Use double bandwidth filters (same center freq)

-ds

Frame GMM computation downsampling ratio

-fdict

word pronunciation dictionary input file

-feat

Feature stream type, depends on the acoustic model

-featparams

containing feature extraction parameters.

-fillprob

Filler word transition probability

-frate

Frame rate

-fsg

format finite state grammar file

-fsgusealtpron

Add alternate pronunciations to FSG

-fsgusefiller

Insert filler words at each state.

-fwdflat

Run forward flat-lexicon search over word lattice (2nd pass)

-fwdflatbeam

Beam width applied to every frame in second-pass flat search

-fwdflatefwid

Minimum number of end frames for a word to be searched in fwdflat search

-fwdflatlw

Language model probability weight for flat lexicon (2nd pass) decoding

-fwdflatsfwin

Window of frames in lattice to search for successor words in fwdflat search

-fwdflatwbeam

Beam width applied to word exits in second-pass flat search

-fwdtree

Run forward lexicon-tree search (1st pass)

-hmm

containing acoustic model files.

-input_endian

Endianness of input data, big or little, ignored if NIST or MS Wav

-jsgf

grammar file

-keyphrase

to spot

-kws

file with keyphrases to spot, one per line

-kws_delay

Delay to wait for best detection score

-kws_plp

Phone loop probability for keyphrase spotting

-kws_threshold

Threshold for p(hyp)/p(alternatives) ratio

-latsize

Initial backpointer table size

-lda

containing transformation matrix to be applied to features (single-stream features only)

-ldadim

Dimensionality of output of feature transformation (0 to use entire matrix)

-lifter

Length of sin-curve for liftering, or 0 for no liftering.

-lm

trigram language model input file

-lmctl

a set of language model

-lmname

language model in -lmctl to use by default

-logbase

Base in which all log-likelihoods calculated

-logfn

to write log messages in

-loglevel

Minimum level of log messages (DEBUG, INFO, WARN, ERROR)

-logspec

Write out logspectral files instead of cepstra

-lowerf

Lower edge of filters

-lpbeam

Beam width applied to last phone in words

-lponlybeam

Beam width applied to last phone in single-phone words

-lw

Language model probability weight

-maxhmmpf

Maximum number of active HMMs to maintain at each frame (or -1 for no pruning)

-maxwpf

Maximum number of distinct word exits at each frame (or -1 for no pruning)

-mdef

definition input file

-mean

gaussian means input file

-mfclogdir

to log feature files to

-min_endfr

Nodes ignored in lattice construction if they persist for fewer than N frames

-mixw

mixture weights input file (uncompressed)

-mixwfloor

Senone mixture weights floor (applied to data from -mixw file)

-mllr

transformation to apply to means and variances

-mmap

Use memory-mapped I/O (if possible) for model files

-ncep

Number of cep coefficients

-nfft

Size of FFT, or 0 to set automatically (recommended)

-nfilt

Number of filter banks

-nwpen

New word transition penalty

-pbeam

Beam width applied to phone transitions

-pip

Phone insertion penalty

-pl_beam

Beam width applied to phone loop search for lookahead

-pl_pbeam

Beam width applied to phone loop transitions for lookahead

-pl_pip

Phone insertion penalty for phone loop

-pl_weight

Weight for phoneme lookahead penalties

-pl_window

Phoneme lookahead window size, in frames

-rawlogdir

to log raw audio files to

-remove_dc

Remove DC offset from each frame

-remove_noise

Remove noise using spectral subtraction

-round_filters

Round mel filter frequencies to DFT points

-samprate

Sampling rate

-seed

Seed for random number generator; if less than zero, pick our own

-sendump

dump (compressed mixture weights) input file

-senlogdir

to log senone score files to

-senmgau

to codebook mapping input file (usually not needed)

-silprob

Silence word transition probability

-smoothspec

Write out cepstral-smoothed logspectral files

-svspec

specification (e.g., 24,0-11/25,12-23/26-38 or 0-12/13-25/26-38)

-tmat

state transition matrix input file

-tmatfloor

HMM state transition probability floor (applied to -tmat file)

-topn

Maximum number of top Gaussians to use in scoring.

-topn_beam

Beam width used to determine top-N Gaussians (or a list, per-feature)

-toprule

rule for JSGF (first public rule is default)

-transform

Which type of transform to use to calculate cepstra (legacy, dct, or htk)

-unit_area

Normalize mel filters to unit area

-upperf

Upper edge of filters

-uw

Unigram weight

-var

gaussian variances input file

-varfloor

Mixture gaussian variance floor (applied to data from -var file)

-varnorm

Variance normalize each utterance (only if CMN == current)

-verbose

Show input filenames

-warp_params

defining the warping function

-warp_type

Warping function type (or shape)

-wbeam

Beam width applied to word exits

-wip

Word insertion penalty

-wlen

Hamming window length

Author

Written by numerous people at CMU from 1994 onwards.  This manual page by David Huggins-Daines <dhdaines@gmail.com>

See Also

pocketsphinx_batch(1), sphinx_fe(1).

Info

2022-09-27