SKP_int SKP_Silk_VAD_FLP(
SKP_Silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */
SKP_Silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */
const SKP_int16 *pIn /* I Input signal */
)
{
SKP_int i, ret, SA_Q8, SNR_dB_Q7, Tilt_Q15;
SKP_int Quality_Bands_Q15[ VAD_N_BANDS ];
ret = SKP_Silk_VAD_GetSA_Q8( &psEnc->sCmn.sVAD, &SA_Q8, &SNR_dB_Q7, Quality_Bands_Q15, &Tilt_Q15,
pIn, psEnc->sCmn.frame_length );
psEnc->speech_activity = ( SKP_float )SA_Q8 / 256.0f;
for( i = 0; i < VAD_N_BANDS; i++ ) {
psEncCtrl->input_quality_bands[ i ] = ( SKP_float )Quality_Bands_Q15[ i ] / 32768.0f;
}
psEncCtrl->input_tilt = ( SKP_float )Tilt_Q15 / 32768.0f;
return ret;
}
int SKP_Silk_encode_frame_FIX(SKP_Silk_encoder_state_FIX * psEnc, /* I/O Pointer to Silk FIX encoder state */
uint8_t * pCode, /* O Pointer to payload */
int16_t * pnBytesOut, /* I/O Pointer to number of payload bytes */
/* input: max length; output: used */
const int16_t * pIn /* I Pointer to input speech frame */
) {
SKP_Silk_encoder_control_FIX sEncCtrl;
int i, nBytes, ret = 0;
int16_t *x_frame, *res_pitch_frame;
int16_t xfw[MAX_FRAME_LENGTH];
int16_t pIn_HP[MAX_FRAME_LENGTH];
int16_t res_pitch[2 * MAX_FRAME_LENGTH + LA_PITCH_MAX];
int LBRR_idx, frame_terminator, SNR_dB_Q7;
const uint16_t *FrameTermination_CDF;
/* Low bitrate redundancy parameters */
uint8_t LBRRpayload[MAX_ARITHM_BYTES];
int16_t nBytesLBRR;
//int32_t Seed[ MAX_LAYERS ];
sEncCtrl.sCmn.Seed = psEnc->sCmn.frameCounter++ & 3;
/**************************************************************/
/* Setup Input Pointers, and insert frame in input buffer */
/*************************************************************/
x_frame = psEnc->x_buf + psEnc->sCmn.frame_length; /* start of frame to encode */
res_pitch_frame = res_pitch + psEnc->sCmn.frame_length; /* start of pitch LPC residual frame */
/****************************/
/* Voice Activity Detection */
/****************************/
ret =
SKP_Silk_VAD_GetSA_Q8(&psEnc->sCmn.sVAD, &psEnc->speech_activity_Q8,
&SNR_dB_Q7, sEncCtrl.input_quality_bands_Q15,
&sEncCtrl.input_tilt_Q15, pIn,
psEnc->sCmn.frame_length);
/*******************************************/
/* High-pass filtering of the input signal */
/*******************************************/
#if HIGH_PASS_INPUT
/* Variable high-pass filter */
SKP_Silk_HP_variable_cutoff_FIX(psEnc, &sEncCtrl, pIn_HP, pIn);
#else
SKP_memcpy(pIn_HP, pIn, psEnc->sCmn.frame_length * sizeof(int16_t));
#endif
#if SWITCH_TRANSITION_FILTERING
/* Ensure smooth bandwidth transitions */
SKP_Silk_LP_variable_cutoff(&psEnc->sCmn.sLP,
x_frame + psEnc->sCmn.la_shape, pIn_HP,
psEnc->sCmn.frame_length);
#else
SKP_memcpy(x_frame + psEnc->sCmn.la_shape, pIn_HP,
psEnc->sCmn.frame_length * sizeof(int16_t));
#endif
/*****************************************/
/* Find pitch lags, initial LPC analysis */
/*****************************************/
SKP_Silk_find_pitch_lags_FIX(psEnc, &sEncCtrl, res_pitch, x_frame);
/************************/
/* Noise shape analysis */
/************************/
SKP_Silk_noise_shape_analysis_FIX(psEnc, &sEncCtrl, res_pitch_frame,
x_frame);
/*****************************************/
/* Prefiltering for noise shaper */
/*****************************************/
SKP_Silk_prefilter_FIX(psEnc, &sEncCtrl, xfw, x_frame);
/***************************************************/
/* Find linear prediction coefficients (LPC + LTP) */
/***************************************************/
SKP_Silk_find_pred_coefs_FIX(psEnc, &sEncCtrl, res_pitch);
/****************************************/
/* Process gains */
/****************************************/
SKP_Silk_process_gains_FIX(psEnc, &sEncCtrl);
psEnc->sCmn.sigtype[psEnc->sCmn.nFramesInPayloadBuf] =
sEncCtrl.sCmn.sigtype;
psEnc->sCmn.QuantOffsetType[psEnc->sCmn.nFramesInPayloadBuf] =
sEncCtrl.sCmn.QuantOffsetType;
/****************************************/
/* Low Bitrate Redundant Encoding */
/****************************************/
nBytesLBRR = MAX_ARITHM_BYTES;
SKP_Silk_LBRR_encode_FIX(psEnc, &sEncCtrl, LBRRpayload, &nBytesLBRR,
xfw);
/*****************************************/
/* Noise shaping quantization */
/*****************************************/
psEnc->NoiseShapingQuantizer(&psEnc->sCmn, &sEncCtrl.sCmn, &psEnc->sNSQ,
xfw,
&psEnc->sCmn.q[psEnc->sCmn.
nFramesInPayloadBuf *
psEnc->sCmn.frame_length],
sEncCtrl.sCmn.NLSFInterpCoef_Q2,
sEncCtrl.PredCoef_Q12[0],
sEncCtrl.LTPCoef_Q14, sEncCtrl.AR2_Q13,
sEncCtrl.HarmShapeGain_Q14,
sEncCtrl.Tilt_Q14, sEncCtrl.LF_shp_Q14,
sEncCtrl.Gains_Q16, sEncCtrl.Lambda_Q10,
sEncCtrl.LTP_scale_Q14);
/**************************************************/
/* Convert speech activity into VAD and DTX flags */
/**************************************************/
if (psEnc->speech_activity_Q8 < SPEECH_ACTIVITY_DTX_THRES_Q8) {
psEnc->sCmn.vadFlag = NO_VOICE_ACTIVITY;
psEnc->sCmn.noSpeechCounter++;
if (psEnc->sCmn.noSpeechCounter > NO_SPEECH_FRAMES_BEFORE_DTX) {
psEnc->sCmn.inDTX = 1;
}
if (psEnc->sCmn.noSpeechCounter > MAX_CONSECUTIVE_DTX) {
psEnc->sCmn.noSpeechCounter = 0;
psEnc->sCmn.inDTX = 0;
}
} else {
psEnc->sCmn.noSpeechCounter = 0;
psEnc->sCmn.inDTX = 0;
psEnc->sCmn.vadFlag = VOICE_ACTIVITY;
}
/****************************************/
/* Initialize arithmetic coder */
/****************************************/
if (psEnc->sCmn.nFramesInPayloadBuf == 0) {
SKP_Silk_range_enc_init(&psEnc->sCmn.sRC);
psEnc->sCmn.nBytesInPayloadBuf = 0;
}
/****************************************/
/* Encode Parameters */
/****************************************/
if (psEnc->sCmn.bitstream_v == BIT_STREAM_V4) {
SKP_Silk_encode_parameters_v4(&psEnc->sCmn, &sEncCtrl.sCmn,
&psEnc->sCmn.sRC);
FrameTermination_CDF = SKP_Silk_FrameTermination_v4_CDF;
} else {
SKP_Silk_encode_parameters(&psEnc->sCmn, &sEncCtrl.sCmn,
&psEnc->sCmn.sRC,
&psEnc->sCmn.q[psEnc->sCmn.
nFramesInPayloadBuf *
psEnc->sCmn.
frame_length]);
FrameTermination_CDF = SKP_Silk_FrameTermination_CDF;
}
/****************************************/
/* Update Buffers and State */
/****************************************/
/* Update Input buffer */
SKP_memmove(psEnc->x_buf, &psEnc->x_buf[psEnc->sCmn.frame_length],
(psEnc->sCmn.frame_length +
psEnc->sCmn.la_shape) * sizeof(int16_t));
/* parameters needed for next frame */
psEnc->sCmn.prev_sigtype = sEncCtrl.sCmn.sigtype;
psEnc->sCmn.prevLag = sEncCtrl.sCmn.pitchL[NB_SUBFR - 1];
psEnc->sCmn.first_frame_after_reset = 0;
if (psEnc->sCmn.sRC.error) {
/* encoder returned error: clear payload buffer */
psEnc->sCmn.nFramesInPayloadBuf = 0;
} else {
psEnc->sCmn.nFramesInPayloadBuf++;
}
/****************************************/
/* finalize payload and copy to output */
/****************************************/
if (psEnc->sCmn.nFramesInPayloadBuf * FRAME_LENGTH_MS >=
psEnc->sCmn.PacketSize_ms) {
LBRR_idx = (psEnc->sCmn.oldest_LBRR_idx + 1) & LBRR_IDX_MASK;
/* Check if FEC information should be added */
frame_terminator = SKP_SILK_LAST_FRAME;
if (psEnc->sCmn.LBRR_buffer[LBRR_idx].usage ==
SKP_SILK_ADD_LBRR_TO_PLUS1) {
frame_terminator = SKP_SILK_LBRR_VER1;
}
if (psEnc->sCmn.LBRR_buffer[psEnc->sCmn.oldest_LBRR_idx].
usage == SKP_SILK_ADD_LBRR_TO_PLUS2) {
frame_terminator = SKP_SILK_LBRR_VER2;
LBRR_idx = psEnc->sCmn.oldest_LBRR_idx;
}
/* Add the frame termination info to stream */
SKP_Silk_range_encoder(&psEnc->sCmn.sRC, frame_terminator,
FrameTermination_CDF);
if (psEnc->sCmn.bitstream_v == BIT_STREAM_V4) {
/* Code excitation signal */
for (i = 0; i < psEnc->sCmn.nFramesInPayloadBuf; i++) {
SKP_Silk_encode_pulses(&psEnc->sCmn.sRC,
psEnc->sCmn.sigtype[i],
psEnc->sCmn.
QuantOffsetType[i],
&psEnc->sCmn.q[i *
psEnc->
sCmn.
frame_length],
psEnc->sCmn.
frame_length);
}
}
/* payload length so far */
SKP_Silk_range_coder_get_length(&psEnc->sCmn.sRC, &nBytes);
/* check that there is enough space in external output buffer, and move data */
if (*pnBytesOut >= nBytes) {
SKP_Silk_range_enc_wrap_up(&psEnc->sCmn.sRC);
SKP_memcpy(pCode, psEnc->sCmn.sRC.buffer,
nBytes * sizeof(uint8_t));
if (frame_terminator > SKP_SILK_MORE_FRAMES &&
*pnBytesOut >=
nBytes + psEnc->sCmn.LBRR_buffer[LBRR_idx].nBytes) {
/* Get old packet and add to payload. */
SKP_memcpy(&pCode[nBytes],
psEnc->sCmn.LBRR_buffer[LBRR_idx].
payload,
psEnc->sCmn.LBRR_buffer[LBRR_idx].
nBytes * sizeof(uint8_t));
nBytes +=
psEnc->sCmn.LBRR_buffer[LBRR_idx].nBytes;
}
*pnBytesOut = nBytes;
/* Update FEC buffer */
SKP_memcpy(psEnc->sCmn.
LBRR_buffer[psEnc->sCmn.oldest_LBRR_idx].
payload, LBRRpayload,
nBytesLBRR * sizeof(uint8_t));
psEnc->sCmn.LBRR_buffer[psEnc->sCmn.oldest_LBRR_idx].
nBytes = nBytesLBRR;
/* This line tells describes how FEC should be used */
psEnc->sCmn.LBRR_buffer[psEnc->sCmn.oldest_LBRR_idx].
usage = sEncCtrl.sCmn.LBRR_usage;
psEnc->sCmn.oldest_LBRR_idx =
(psEnc->sCmn.oldest_LBRR_idx + 1) & LBRR_IDX_MASK;
/* Reset number of frames in payload buffer */
psEnc->sCmn.nFramesInPayloadBuf = 0;
} else {
/* Not enough space: Payload will be discarded */
*pnBytesOut = 0;
nBytes = 0;
psEnc->sCmn.nFramesInPayloadBuf = 0;
ret = SKP_SILK_ENC_PAYLOAD_BUF_TOO_SHORT;
}
} else {
/* no payload for you this time */
*pnBytesOut = 0;
/* Encode that more frames follows */
frame_terminator = SKP_SILK_MORE_FRAMES;
SKP_Silk_range_encoder(&psEnc->sCmn.sRC, frame_terminator,
FrameTermination_CDF);
/* payload length so far */
SKP_Silk_range_coder_get_length(&psEnc->sCmn.sRC, &nBytes);
if (psEnc->sCmn.bitstream_v == BIT_STREAM_V4) {
/* Take into account the q signal that isnt in the bitstream yet */
nBytes += SKP_Silk_pulses_to_bytes(&psEnc->sCmn,
&psEnc->sCmn.
q[(psEnc->sCmn.
nFramesInPayloadBuf
-
1) *
psEnc->sCmn.
frame_length]);
}
}
/* Check for arithmetic coder errors */
if (psEnc->sCmn.sRC.error) {
ret = SKP_SILK_ENC_INTERNAL_ERROR;
}
/* simulate number of ms buffered in channel because of exceeding TargetRate */
assert((8 * 1000 *
((int64_t) nBytes -
(int64_t) psEnc->sCmn.nBytesInPayloadBuf)) ==
SKP_SAT32(8 * 1000 *
((int64_t) nBytes -
(int64_t) psEnc->sCmn.nBytesInPayloadBuf)));
assert(psEnc->sCmn.TargetRate_bps > 0);
psEnc->BufferedInChannel_ms +=
SKP_DIV32(8 * 1000 * (nBytes - psEnc->sCmn.nBytesInPayloadBuf),
psEnc->sCmn.TargetRate_bps);
psEnc->BufferedInChannel_ms -= FRAME_LENGTH_MS;
psEnc->BufferedInChannel_ms =
SKP_LIMIT(psEnc->BufferedInChannel_ms, 0, 100);
psEnc->sCmn.nBytesInPayloadBuf = nBytes;
if (psEnc->speech_activity_Q8 > WB_DETECT_ACTIVE_SPEECH_LEVEL_THRES_Q8) {
psEnc->sCmn.sSWBdetect.ActiveSpeech_ms =
SKP_ADD_POS_SAT32(psEnc->sCmn.sSWBdetect.ActiveSpeech_ms,
FRAME_LENGTH_MS);
}
return (ret);
}
