ref: 10f28719ce4e3efef7b237fa2f8a339c50a44b7b
parent: 48102ce56e922331446cc7a1c0833ecbb1e6532c
author: cbagwell <cbagwell>
date: Sun Mar 18 21:46:30 EDT 2007
Adding -l option to silence to leave in a minium amount of silence. Useful for removing long pauses between words but not remove the pause all together.
--- a/ChangeLog
+++ b/ChangeLog
@@ -20,6 +20,8 @@
o Show (with --octave) compand transfer function. (robs)
o Allow e.g. "vol 6dB" (as well as "vol 6 dB"). (robs)
o Changed deemph to be a true biquad for better accuracy. (robs)
+ o Add option to silence effect to leave periods of silence
+ in and only strip out extra silence. Mark Schreiber
Other new features:
--- a/sox.1
+++ b/sox.1
@@ -1856,7 +1856,7 @@
Reverse the audio completely.
Requires disk space to store the data to be reversed.
.TP
-\fBsilence \fIabove-periods\fR [\fIduration threshold\fR[\fBd\fR\^|\^\fB%\fR] [\fIbelow-periods duration threshold\fR[\fBd\fR\^|\^\fB%\fR]]
+\fBsilence [\fB\-l\fR] \fIabove-periods\fR [\fIduration threshold\fR[\fBd\fR\^|\^\fB%\fR] [\fIbelow-periods duration threshold\fR[\fBd\fR\^|\^\fB%\fR]]
.SP
Removes silence from the beginning, middle, or end of the audio. Silence is anything below a specified threshold.
.SP
@@ -1909,6 +1909,13 @@
effect should restart processing as specified by the
\fIabove-periods\fR, making it suitable for removing periods of
silence in the middle of the audio.
+.SP
+The option
+.B \-l
+indicates that \fIbelow-periods\fR \fIduration\fR length of audio
+should be left intact at the beginning of each period of silence.
+For example, if you want to remove long pauses between words
+but do not want to remove the pauses completely.
.SP
The \fIperiod\fR counts are in units of samples. \fIDuration\fR counts may be in the format of hh:mm:ss.frac, or the exact count of samples. \fIThreshold\fR numbers may be suffixed with
.B d
--- a/src/silence.c
+++ b/src/silence.c
@@ -61,6 +61,8 @@
sox_size_t window_size;
double rms_sum;
+ char leave_silence;
+
/* State Machine */
char mode;
} *silence_t;
@@ -83,6 +85,16 @@
silence_t silence = (silence_t) effp->priv;
int parse_count;
+ /* check for option switches */
+ silence->leave_silence = sox_false;
+ if (n > 0)
+ {+ if (!strcmp("-l", *argv)) {+ n--; argv++;
+ silence->leave_silence = sox_true;
+ }
+ }
+
if (n < 1)
{sox_fail(sox_silence_effect.usage);
@@ -457,21 +469,43 @@
break;
case SILENCE_COPY:
- /* Attempts to copy samples into output buffer. If not
- * looking for silence to terminate copy then blindly
- * copy data into output buffer.
+ /* Attempts to copy samples into output buffer.
*
- * If looking for silence, then see if input sample is above
- * threshold. If found then flush out hold off buffer
- * and copy over to output buffer. Tell user about
- * input and output processing.
+ * Case B:
+ * If not looking for silence to terminate copy then
+ * blindly copy data into output buffer.
*
- * If not above threshold then store in hold off buffer
- * and do not write to output buffer. Tell user input
- * was processed.
+ * Case A:
*
- * If hold off buffer is full then stop copying data and
- * discard data in hold off buffer.
+ * Case 1a:
+ * If previous silence was detect then see if input sample is
+ * above threshold. If found then flush out hold off buffer
+ * and copy over to output buffer.
+ *
+ * Case 1b:
+ * If no previous silence detect then see if input sample
+ * is above threshold. If found then copy directly
+ * to output buffer.
+ *
+ * Case 2:
+ * If not above threshold then silence is detect so
+ * store in hold off buffer and do not write to output
+ * buffer. Even though it wasn't put in output
+ * buffer, inform user that input was consumed.
+ *
+ * If hold off buffer is full after this then stop
+ * copying data and discard data in hold off buffer.
+ *
+ * Special leave_silence logic:
+ *
+ * During this mode, go ahead and copy input
+ * samples to output buffer instead of holdoff buffer
+ * Then also short ciruit any flushes that would occur
+ * when non-silence is detect since samples were already
+ * copied. This has the effect of always leaving
+ * holdoff[] amount of silence but deleting any
+ * beyond that amount.
+ *
*/
silence_copy:
nrOfTicks = min((*isamp-nrOfInSamplesRead),
@@ -479,6 +513,7 @@
effp->ininfo.channels;
if (silence->stop)
{+ /* Case A */
for(i = 0; i < nrOfTicks; i++)
{threshold = 1;
@@ -490,16 +525,24 @@
silence->stop_unit);
}
- /* If above threshold, check to see if we where holding
+ /* Case 1a
+ * If above threshold, check to see if we where holding
* off previously. If so then flush this buffer.
* We haven't incremented any pointers yet so nothing
* is lost.
+ *
+ * If user wants to leave_silence, then we
+ * were already copying the data and so no
+ * need to flush the old data. Just resume
+ * copying as if we were not holding off.
*/
- if (threshold && silence->stop_holdoff_end)
+ if (threshold && silence->stop_holdoff_end
+ && !silence->leave_silence)
{silence->mode = SILENCE_COPY_FLUSH;
goto silence_copy_flush;
}
+ /* Case 1b */
else if (threshold)
{/* Not holding off so copy into output buffer */
@@ -511,6 +554,7 @@
nrOfOutSamplesWritten++;
}
}
+ /* Case 2 */
else if (!threshold)
{/* Add to holdoff buffer */
@@ -517,6 +561,10 @@
for (j = 0; j < effp->ininfo.channels; j++)
{update_rms(effp, *ibuf);
+ if (silence->leave_silence) {+ *obuf++ = *ibuf;
+ nrOfOutSamplesWritten++;
+ }
silence->stop_holdoff[
silence->stop_holdoff_end++] = *ibuf++;
nrOfInSamplesRead++;
@@ -570,6 +618,7 @@
} /* Trimming off backend */
else /* !(silence->stop) */
{+ /* Case B */
memcpy(obuf, ibuf, sizeof(sox_sample_t)*nrOfTicks*
effp->ininfo.channels);
nrOfInSamplesRead += (nrOfTicks*effp->ininfo.channels);