speex_codec.c

一个开源的sip源代码

    attr->setting.frm_per_pkt = 1;

    /* Default flags. */
    attr->setting.cng = 1;
    attr->setting.plc = 1;
    attr->setting.penh =1 ;
    attr->setting.vad = 1;

    return PJ_SUCCESS;
}

/*
 * Enum codecs supported by this factory (i.e. only Speex!).
 */
static pj_status_t spx_enum_codecs(pjmedia_codec_factory *factory, 
				    unsigned *count, 
				    pjmedia_codec_info codecs[])
{
    unsigned max;
    int i;  /* Must be signed */

    PJ_UNUSED_ARG(factory);
    PJ_ASSERT_RETURN(codecs && *count > 0, PJ_EINVAL);

    max = *count;
    *count = 0;

    /*
     * We return three codecs here, and in this order:
     *	- ultra-wideband, wideband, and narrowband.
     */
    for (i=PJ_ARRAY_SIZE(spx_factory.speex_param)-1; i>=0 && *count<max; --i) {

	if (!spx_factory.speex_param[i].enabled)
	    continue;

	pj_bzero(&codecs[*count], sizeof(pjmedia_codec_info));
	codecs[*count].encoding_name = pj_str("speex");
	codecs[*count].pt = spx_factory.speex_param[i].pt;
	codecs[*count].type = PJMEDIA_TYPE_AUDIO;
	codecs[*count].clock_rate = spx_factory.speex_param[i].clock_rate;
	codecs[*count].channel_cnt = 1;

	++*count;
    }

    return PJ_SUCCESS;
}

/*
 * Allocate a new Speex codec instance.
 */
static pj_status_t spx_alloc_codec( pjmedia_codec_factory *factory, 
				    const pjmedia_codec_info *id,
				    pjmedia_codec **p_codec)
{
    pjmedia_codec *codec;
    struct spx_private *spx;

    PJ_ASSERT_RETURN(factory && id && p_codec, PJ_EINVAL);
    PJ_ASSERT_RETURN(factory == &spx_factory.base, PJ_EINVAL);


    pj_mutex_lock(spx_factory.mutex);

    /* Get free nodes, if any. */
    if (!pj_list_empty(&spx_factory.codec_list)) {
	codec = spx_factory.codec_list.next;
	pj_list_erase(codec);
    } else {
	codec = PJ_POOL_ZALLOC_T(spx_factory.pool, pjmedia_codec);
	PJ_ASSERT_RETURN(codec != NULL, PJ_ENOMEM);
	codec->op = &spx_op;
	codec->factory = factory;
	codec->codec_data = pj_pool_alloc(spx_factory.pool,
					  sizeof(struct spx_private));
    }

    pj_mutex_unlock(spx_factory.mutex);

    spx = (struct spx_private*) codec->codec_data;
    spx->enc = NULL;
    spx->dec = NULL;

    if (id->clock_rate <= 8000)
	spx->param_id = PARAM_NB;
    else if (id->clock_rate <= 16000)
	spx->param_id = PARAM_WB;
    else
	spx->param_id = PARAM_UWB;

    *p_codec = codec;
    return PJ_SUCCESS;
}

/*
 * Free codec.
 */
static pj_status_t spx_dealloc_codec( pjmedia_codec_factory *factory, 
				      pjmedia_codec *codec )
{
    struct spx_private *spx;

    PJ_ASSERT_RETURN(factory && codec, PJ_EINVAL);
    PJ_ASSERT_RETURN(factory == &spx_factory.base, PJ_EINVAL);

    /* Close codec, if it's not closed. */
    spx = (struct spx_private*) codec->codec_data;
    if (spx->enc != NULL || spx->dec != NULL) {
	spx_codec_close(codec);
    }

    /* Put in the free list. */
    pj_mutex_lock(spx_factory.mutex);
    pj_list_push_front(&spx_factory.codec_list, codec);
    pj_mutex_unlock(spx_factory.mutex);

    return PJ_SUCCESS;
}

/*
 * Init codec.
 */
static pj_status_t spx_codec_init( pjmedia_codec *codec, 
				   pj_pool_t *pool )
{
    PJ_UNUSED_ARG(codec);
    PJ_UNUSED_ARG(pool);
    return PJ_SUCCESS;
}

/*
 * Open codec.
 */
static pj_status_t spx_codec_open( pjmedia_codec *codec, 
				   pjmedia_codec_param *attr )
{
    struct spx_private *spx;
    int id, tmp;

    spx = (struct spx_private*) codec->codec_data;
    id = spx->param_id;

    /* Only supports one frame per packet */
    PJ_ASSERT_RETURN(attr->setting.frm_per_pkt==1, PJ_EINVAL);

    /* 
     * Create and initialize encoder. 
     */
    spx->enc = speex_encoder_init(spx_factory.speex_param[id].mode);
    if (!spx->enc)
	return PJMEDIA_CODEC_EFAILED;
    speex_bits_init(&spx->enc_bits);

    /* Set the quality*/
    if (spx_factory.speex_param[id].quality != -1) {
	speex_encoder_ctl(spx->enc, SPEEX_SET_QUALITY, 
			  &spx_factory.speex_param[id].quality);
    }

    /* Sampling rate. */
    tmp = attr->info.clock_rate;
    speex_encoder_ctl(spx->enc, SPEEX_SET_SAMPLING_RATE, 
		      &spx_factory.speex_param[id].clock_rate);

    /* VAD */
    tmp = (attr->setting.vad != 0);
    speex_encoder_ctl(spx->enc, SPEEX_SET_VAD, &tmp);
    speex_encoder_ctl(spx->enc, SPEEX_SET_DTX, &tmp);

    /* Complexity */
    if (spx_factory.speex_param[id].complexity != -1) {
	speex_encoder_ctl(spx->enc, SPEEX_SET_COMPLEXITY, 
			  &spx_factory.speex_param[id].complexity);
    }

    /* 
     * Create and initialize decoder. 
     */
    spx->dec = speex_decoder_init(spx_factory.speex_param[id].mode);
    if (!spx->dec) {
	spx_codec_close(codec);
	return PJMEDIA_CODEC_EFAILED;
    }
    speex_bits_init(&spx->dec_bits);

    /* Sampling rate. */
    speex_decoder_ctl(spx->dec, SPEEX_SET_SAMPLING_RATE, 
		      &spx_factory.speex_param[id].clock_rate);

    /* PENH */
    tmp = attr->setting.penh;
    speex_decoder_ctl(spx->dec, SPEEX_SET_ENH, &tmp);

    return PJ_SUCCESS;
}

/*
 * Close codec.
 */
static pj_status_t spx_codec_close( pjmedia_codec *codec )
{
    struct spx_private *spx;

    spx = (struct spx_private*) codec->codec_data;

    /* Destroy encoder*/
    if (spx->enc) {
	speex_encoder_destroy( spx->enc );
	spx->enc = NULL;
	speex_bits_destroy( &spx->enc_bits );
    }

    /* Destroy decoder */
    if (spx->dec) {
	speex_decoder_destroy( spx->dec);
	spx->dec = NULL;
	speex_bits_destroy( &spx->dec_bits );
    }

    return PJ_SUCCESS;
}


/*
 * Modify codec settings.
 */
static pj_status_t  spx_codec_modify(pjmedia_codec *codec, 
				     const pjmedia_codec_param *attr )
{
    struct spx_private *spx;
    int tmp;

    spx = (struct spx_private*) codec->codec_data;

    /* Only supports one frame per packet */
    PJ_ASSERT_RETURN(attr->setting.frm_per_pkt==1, PJ_EINVAL);

    /* VAD */
    tmp = (attr->setting.vad != 0);
    speex_encoder_ctl(spx->enc, SPEEX_SET_VAD, &tmp);
    speex_encoder_ctl(spx->enc, SPEEX_SET_DTX, &tmp);

    /* PENH */
    tmp = attr->setting.penh;
    speex_decoder_ctl(spx->dec, SPEEX_SET_ENH, &tmp);

    return PJ_SUCCESS;
}


/*
 * Get frames in the packet.
 */
static pj_status_t  spx_codec_parse( pjmedia_codec *codec,
				     void *pkt,
				     pj_size_t pkt_size,
				     const pj_timestamp *ts,
				     unsigned *frame_cnt,
				     pjmedia_frame frames[])
{
    struct spx_private *spx;
    unsigned frame_size, samples_per_frame;
    unsigned count;

    spx = (struct spx_private*) codec->codec_data;

    frame_size = spx_factory.speex_param[spx->param_id].framesize;
    samples_per_frame = spx_factory.speex_param[spx->param_id].samples_per_frame;

    /* Don't really know how to do this... */
    count = 0;
    while (pkt_size >= frame_size && count < *frame_cnt) {
	frames[count].buf = pkt;
	frames[count].size = frame_size;
	frames[count].type = PJMEDIA_FRAME_TYPE_AUDIO;
	frames[count].timestamp.u64 = ts->u64 + count * samples_per_frame;

	pkt_size -= frame_size;
	++count;
	pkt = ((char*)pkt) + frame_size;
    }

    /* Just in case speex has silence frame which size is less than normal
     * frame size...
     */
    if (pkt_size && count < *frame_cnt) {
	frames[count].buf = pkt;
	frames[count].size = pkt_size;
	frames[count].type = PJMEDIA_FRAME_TYPE_AUDIO;
	frames[count].timestamp.u64 = ts->u64 + count * samples_per_frame;
	++count;
    }

    *frame_cnt = count;
    return PJ_SUCCESS;
}

/*
 * Encode frame.
 */
static pj_status_t spx_codec_encode( pjmedia_codec *codec, 
				     const struct pjmedia_frame *input,
				     unsigned output_buf_len, 
				     struct pjmedia_frame *output)
{
    struct spx_private *spx;
    unsigned sz;
    int tx;

    spx = (struct spx_private*) codec->codec_data;

    if (input->type != PJMEDIA_FRAME_TYPE_AUDIO) {
	output->size = 0;
	output->buf = NULL;
	output->timestamp = input->timestamp;
	output->type = input->type;
	return PJ_SUCCESS;
    }

    /* Flush all the bits in the struct so we can encode a new frame */
    speex_bits_reset(&spx->enc_bits);

    /* Encode the frame */
    tx = speex_encode_int(spx->enc, (spx_int16_t*)input->buf, 
			  &spx->enc_bits);

    /* Check if we need not to transmit the frame (DTX) */
    if (tx == 0) {
	output->buf = NULL;
	output->size = 0;
	output->timestamp.u64 = input->timestamp.u64;
	output->type = PJMEDIA_FRAME_TYPE_NONE;
	return PJ_SUCCESS;
    }

    /* Check size. */
    sz = speex_bits_nbytes(&spx->enc_bits);
    pj_assert(sz <= output_buf_len);

    /* Copy the bits to an array of char that can be written */
    output->size = speex_bits_write(&spx->enc_bits, 
				    (char*)output->buf, output_buf_len);
    output->type = PJMEDIA_FRAME_TYPE_AUDIO;
    output->timestamp = input->timestamp;

    return PJ_SUCCESS;
}

/*
 * Decode frame.
 */
static pj_status_t spx_codec_decode( pjmedia_codec *codec, 
				     const struct pjmedia_frame *input,
				     unsigned output_buf_len, 
				     struct pjmedia_frame *output)
{
    struct spx_private *spx;

    spx = (struct spx_private*) codec->codec_data;

    if (input->type != PJMEDIA_FRAME_TYPE_AUDIO) {
	pj_bzero(output->buf, output_buf_len);
	output->size = 320;
	output->timestamp.u64 = input->timestamp.u64;
	output->type = PJMEDIA_FRAME_TYPE_AUDIO;
	return PJ_SUCCESS;
    }

    /* Copy the data into the bit-stream struct */
    speex_bits_read_from(&spx->dec_bits, (char*)input->buf, input->size);

    /* Decode the data */
    speex_decode_int(spx->dec, &spx->dec_bits, (spx_int16_t*)output->buf);

    output->type = PJMEDIA_FRAME_TYPE_AUDIO;
    output->size = speex_bits_nbytes(&spx->dec_bits);
    pj_assert(output->size <= (unsigned)output_buf_len);
    output->timestamp.u64 = input->timestamp.u64;


    return PJ_SUCCESS;
}

/* 
 * Recover lost frame.
 */
static pj_status_t  spx_codec_recover(pjmedia_codec *codec, 
				      unsigned output_buf_len, 
				      struct pjmedia_frame *output)
{
    struct spx_private *spx;
    unsigned count;

    /* output_buf_len is unreferenced when building in Release mode */
    PJ_UNUSED_ARG(output_buf_len);

    spx = (struct spx_private*) codec->codec_data;

    count = spx_factory.speex_param[spx->param_id].clock_rate * 20 / 1000;
    pj_assert(count <= output_buf_len/2);

    /* Recover packet loss */
    speex_decode_int(spx->dec, NULL, (spx_int16_t*) output->buf);

    output->size = count * 2;

    return PJ_SUCCESS;
}


#endif	/* PJMEDIA_HAS_SPEEX_CODEC */