The filter
module can filter out the following streams:
The filtering operation can be based on either broad tags, such as speech or silence, or on a few specific word labels.
The filter
module has multiple modes of operation.
Mode | Input | Output | Description |
---|---|---|---|
SPEECH
|
w
|
w
|
Accepts a word stream, and produces only the tokens that are not silence. |
EXCLUSIVE
|
w
|
w
|
Accepts a word stream, and produces only those tokens that do not match the specified filters. |
INCLUSIVE
|
w
|
w
|
Accepts a word stream, and produces only those tokens that match the specified filters. |
FEAT_SPEECH
|
nf,
|
nf
|
Accepts a word stream (or language identification results stream) and normalized frame stream, and produces only frames within segments that are not labeled as silence. |
FEAT_EXCLUSIVE
|
nf, w
|
nf
|
Accepts a word label stream and normalized frame stream and produces only frames within segments for which the word does not match the specified filters. |
FEAT_INCLUSIVE
|
nf,
|
nf
|
Accepts a word label stream and a normalized frame stream, and produces only frames within segments for which the word matches the specified filters. |
FEAT_EXCLUSIVE
|
nf, lid
|
nf
|
Accepts a language identification results stream and a normalized frame stream, and produces only frames within segments for which the word does not match the specified filters. |
FEAT_INCLUSIVE
|
nf, lid
|
nf
|
Accepts a language identification results stream and a normalized frame stream, and produces only frames within segments for which the word matches the specified filters. |
LF_INCLUSIVE
|
lf, w
|
lf
|
Accepts a word label stream and a language ID feature stream, and outputs only those language ID features which are within segments for which the word matches the specified filters. |
LF_EXCLUSIVE
|
lf, w
|
lf
|
Accepts a word label stream and a language ID feature stream, and outputs only those language ID features which are not within segments for which the word matches the specified filters. |
WAV_SPEECH
|
a
|
a
|
Accepts an audio stream and produces only audio samples within segments that are not labeled as silence. |
WAV_EXCLUSIVE
|
a,
|
a , |
Accepts a word label stream and an audio stream, and produces only audio samples within segments for which the label does not match the specified filters. |
WAV_INCLUSIVE
|
a,
|
a
|
Accepts a word label stream and an audio stream, and produces only audio samples within segments for which the label matches the specified filters. |
WAV_EXCLUSIVE
|
a,
|
a
|
Accepts a language identification results stream and an audio stream, and produces only audio samples within segments for which the label does not match the specified filters. |
WAV_INCLUSIVE
|
a,
|
a
|
Accepts a language identification results stream and an audio stream, and produces only audio samples within segments for which the label matches the specified filters. |
Examples:
w2 <- filter (SPEECH, w1)
w2 <- filter (EXCLUSIVE, w1)
w2 <- filter (INCLUSIVE, w1)
nf2 <- filter (FEAT_SPEECH, nf1, w)
nf2 <- filter (FEAT_EXCLUSIVE, nf1, w)
nf2 <- filter (FEAT_EXCLUSIVE, nf1, lid)
a2 <- filter (FEAT_INCLUSIVE, a1 w)
a2 <- filter (FEAT_INCLUSIVE, a1, lid)
|