[alsa-devel] ALSA 1.0.20 Speex PCM Plugin

[Takashi Iwai]

At Tue, 2 Jun 2009 16:39:46 -0400,
Robert Krakora wrote:
> 
> Hello,
> 
> Has anyone successfully employed the ALSA 1.0.20 Speex PCM Plugin?  I
> followed the "speexdsp.txt" document under the 'doc' directory but the
> result was the following error:
> 
> [root at vizioroom105 ~]# arecord -Dplug:mic poopy.wav
> Recording WAVE 'poopy.wav' : Unsigned 8 bit, Rate 8000 Hz, Mono
> ALSA lib pcm_params.c:2135:(snd1_pcm_hw_refine_slave) Slave PCM not usable
> arecord: set_params:957: Broken configuration for this PCM: no
> configurations available

It's because the slave plugin of pcm.mic is "hw" and that doesn't
support mono streams but only stereo.  speex plugin requires a mono
stream explicitly.

Wrap speex plugin over the default or add plug layer inside it, too.

BTW, I found that I didn't put any echo-cancelling code in the speex
plugin.  It was just written for denoising.

The below is a quick hack to add the echo-cancelling part.  Give it a try
(although it's totally untested :)


Takashi

---
diff --git a/doc/speexdsp.txt b/doc/speexdsp.txt
index 875fc19..1937de6 100644
--- a/doc/speexdsp.txt
+++ b/doc/speexdsp.txt
@@ -12,7 +12,7 @@ using libspeex DSP API.  You can use the plugin with the plugin type
 
 Then record like
 
-	% arecord -fdat -c1 -Dplug:speex foo.wav
+	% arecord -fdat -c1 -Dplug:my_pcm foo.wav
 
 so that you'll get 48kHz mono stream with the denoising effect.
 
@@ -44,6 +44,16 @@ The following parameters can be set optionally:
 
   A boolean value to enable/disable dereverb function.  Default is no.
 
+* echo
+
+  A boolean value to enable/disable echo-cancellation function.
+  Default is no.
+
+* filter_length
+
+  Number of samples of echo to cancel.  As default it's 256.
+
+
 For example, you can enable agc like
 
 	pcm.my_pcm {
diff --git a/speex/pcm_speex.c b/speex/pcm_speex.c
index 7bb9213..38b3582 100644
--- a/speex/pcm_speex.c
+++ b/speex/pcm_speex.c
@@ -1,5 +1,5 @@
 /*
- * Speex preprocess plugin
+ * Speex DSP plugin
  *
  * Copyright (c) 2009 by Takashi Iwai <tiwai at suse.de>
  *
@@ -21,12 +21,15 @@
 #include <alsa/asoundlib.h>
 #include <alsa/pcm_external.h>
 #include <speex/speex_preprocess.h>
+#include <speex/speex_echo.h>
 
-/* preprocessing parameters */
+/* DSP parameters */
 struct spx_parms {
 	int frames;
 	int denoise;
 	int agc;
+	int echo;
+	int filter_length;
 	float agc_level;
 	int dereverb;
 	float dereverb_decay;
@@ -38,7 +41,9 @@ typedef struct {
 	struct spx_parms parms;
 	/* instance and intermedate buffer */
 	SpeexPreprocessState *state;
+	SpeexEchoState *echo_state;
 	short *buf;
+	short *outbuf;
 	/* running states */
 	unsigned int filled;
 	unsigned int processed;
@@ -64,6 +69,12 @@ spx_transfer(snd_pcm_extplug_t *ext,
 	short *src = area_addr(src_areas, src_offset);
 	short *dst = area_addr(dst_areas, dst_offset);
 	unsigned int count = size;
+	short *databuf;
+
+	if (spx->parms.echo)
+		databuf = spx->outbuf;
+	else
+		databuf = spx->buf;
 
 	while (count > 0) {
 		unsigned int chunk;
@@ -72,14 +83,19 @@ spx_transfer(snd_pcm_extplug_t *ext,
 		else
 			chunk = count;
 		if (spx->processed)
-			memcpy(dst, spx->buf + spx->filled, chunk * 2);
+			memcpy(dst, databuf + spx->filled, chunk * 2);
 		else
 			memset(dst, 0, chunk * 2);
 		dst += chunk;
 		memcpy(spx->buf + spx->filled, src, chunk * 2);
 		spx->filled += chunk;
 		if (spx->filled == spx->parms.frames) {
-			speex_preprocess_run(spx->state, spx->buf);
+			if (spx->parms.echo)
+				speex_echo_capture(spx->echo_state, spx->buf,
+						   spx->outbuf);
+			speex_preprocess_run(spx->state, databuf);
+			if (spx->parms.echo)
+				speex_echo_playback(spx->echo_state, databuf);
 			spx->processed = 1;
 			spx->filled = 0;
 		}
@@ -101,13 +117,34 @@ static int spx_init(snd_pcm_extplug_t *ext)
 	}
 	memset(spx->buf, 0, spx->parms.frames * 2);
 
-	if (spx->state)
+	if (spx->state) {
 		speex_preprocess_state_destroy(spx->state);
+		spx->state = NULL;
+	}
+	if (spx->echo_state) {
+		speex_echo_state_destroy(spx->echo_state);
+		spx->echo_state = NULL;
+	}
+
+	if (spx->parms.echo) {
+		spx->echo_state = speex_echo_state_init(spx->parms.frames,
+						spx->parms.filter_length);
+		if (!spx->echo_state)
+			return -EIO;
+		speex_echo_ctl(spx->echo_state, SPEEX_ECHO_SET_SAMPLING_RATE,
+			       &spx->ext.rate);
+	}
+
 	spx->state = speex_preprocess_state_init(spx->parms.frames,
 						 spx->ext.rate);
 	if (!spx->state)
 		return -EIO;
 
+	if (spx->parms.echo)
+		speex_preprocess_ctl(spx->state,
+				     SPEEX_PREPROCESS_SET_ECHO_STATE,
+				     spx->echo_state);
+
 	speex_preprocess_ctl(spx->state, SPEEX_PREPROCESS_SET_DENOISE,
 			     &spx->parms.denoise);
 	speex_preprocess_ctl(spx->state, SPEEX_PREPROCESS_SET_AGC,
@@ -132,6 +169,8 @@ static int spx_close(snd_pcm_extplug_t *ext)
 	free(spx->buf);
 	if (spx->state)
 		speex_preprocess_state_destroy(spx->state);
+	if (spx->echo_state)
+		speex_echo_state_destroy(spx->echo_state);
 	return 0;
 }
 
@@ -205,6 +244,8 @@ SND_PCM_PLUGIN_DEFINE_FUNC(speex)
 		.dereverb = 0,
 		.dereverb_decay = 0,
 		.dereverb_level = 0,
+		.echo = 0,
+		.filter_length = 256,
 	};
 
 	snd_config_for_each(i, next, conf) {
@@ -242,6 +283,12 @@ SND_PCM_PLUGIN_DEFINE_FUNC(speex)
 				     &parms.dereverb_level);
 		if (err)
 			goto ok;
+		err = get_bool_parm(n, id, "echo", &parms.echo);
+		if (err)
+			goto ok;
+		err = get_int_parm(n, id, "filter_length", &parms.filter_length);
+		if (err)
+			goto ok;
 		SNDERR("Unknown field %s", id);
 		err = -EINVAL;
 	ok:
@@ -259,7 +306,7 @@ SND_PCM_PLUGIN_DEFINE_FUNC(speex)
 		return -ENOMEM;
 
 	spx->ext.version = SND_PCM_EXTPLUG_VERSION;
-	spx->ext.name = "Speex Denoise Plugin";
+	spx->ext.name = "Speex DSP Plugin";
 	spx->ext.callback = &speex_callback;
 	spx->ext.private_data = spx;
 	spx->parms = parms;