diff --git a/dnn/lpcnet_plc.c b/dnn/lpcnet_plc.c index 761c907517b06a126bf0c1e0de0a427aba77b7fa..0a1a95aaaef0ee8c7013ee33f97b22cb1c0c2d56 100644 --- a/dnn/lpcnet_plc.c +++ b/dnn/lpcnet_plc.c @@ -154,11 +154,13 @@ static int get_fec_or_pred(LPCNetPLCState *st, float *out) { } } -static void fec_rewind(LPCNetPLCState *st, int offset) { - st->fec_read_pos -= offset; - if (st->fec_read_pos < st->fec_keep_pos) { - st->fec_read_pos = st->fec_keep_pos; - } +static void queue_features(LPCNetPLCState *st, const float *features) { + OPUS_MOVE(&st->cont_features[0], &st->cont_features[NB_FEATURES], (CONT_VECTORS-1)*NB_FEATURES); + OPUS_COPY(&st->cont_features[(CONT_VECTORS-1)*NB_FEATURES], features, NB_FEATURES); +} + +static void replace_features(LPCNetPLCState *st, const float *features) { + OPUS_COPY(&st->cont_features[(CONT_VECTORS-1)*NB_FEATURES], features, NB_FEATURES); } /* In this causal version of the code, the DNN model implemented by compute_plc_pred() @@ -170,29 +172,18 @@ int lpcnet_plc_update(LPCNetPLCState *st, opus_int16 *pcm) { float plc_features[2*NB_BANDS+NB_FEATURES+1]; for (i=0;i<FRAME_SIZE;i++) x[i] = pcm[i]; burg_cepstral_analysis(plc_features, x); + lpcnet_compute_single_frame_features_float(&st->enc, x, st->features); if (st->blend) { + replace_features(st, st->features); if (FEATURES_DELAY > 0) st->plc_net = st->plc_copy[FEATURES_DELAY-1]; - fec_rewind(st, FEATURES_DELAY); - } - /* Update state. */ - /*fprintf(stderr, "update state\n");*/ - for (i=0;i<FRAME_SIZE;i++) x[i] = pcm[i]; - preemphasis(x, &st->enc.mem_preemph, x, PREEMPHASIS, FRAME_SIZE); - compute_frame_features(&st->enc, x); - process_single_frame(&st->enc, NULL); - if (!st->blend) { - OPUS_COPY(&plc_features[2*NB_BANDS], st->enc.features, NB_FEATURES); + } else { + queue_features(st, st->features); + OPUS_COPY(&plc_features[2*NB_BANDS], st->features, NB_FEATURES); plc_features[2*NB_BANDS+NB_FEATURES] = 1; compute_plc_pred(st, st->features, plc_features); - /* Discard an FEC frame that we know we will no longer need. */ - if (st->fec_skip) st->fec_skip--; - else if (st->fec_read_pos < st->fec_fill_pos) st->fec_read_pos++; - st->fec_keep_pos = IMAX(0, IMAX(st->fec_keep_pos, st->fec_read_pos-FEATURES_DELAY-1)); - OPUS_MOVE(&st->cont_features[0], &st->cont_features[NB_FEATURES], (CONT_VECTORS-1)*NB_FEATURES); } - OPUS_COPY(&st->cont_features[(CONT_VECTORS-1)*NB_FEATURES], st->enc.features, NB_FEATURES); - OPUS_MOVE(st->pcm, &st->pcm[FRAME_SIZE], FARGAN_CONT_SAMPLES-FRAME_SIZE); - for (i=0;i<FRAME_SIZE;i++) st->pcm[FARGAN_CONT_SAMPLES-FRAME_SIZE+i] = (1.f/32768.f)*pcm[i]; + OPUS_MOVE(st->pcm, &st->pcm[FRAME_SIZE], PLC_BUF_SIZE-FRAME_SIZE); + for (i=0;i<FRAME_SIZE;i++) st->pcm[PLC_BUF_SIZE-FRAME_SIZE+i] = (1.f/32768.f)*pcm[i]; st->loss_count = 0; st->blend = 0; return 0; @@ -203,12 +194,10 @@ int lpcnet_plc_conceal(LPCNetPLCState *st, opus_int16 *pcm) { int i; if (st->blend == 0) { get_fec_or_pred(st, st->features); - OPUS_MOVE(&st->cont_features[0], &st->cont_features[NB_FEATURES], (CONT_VECTORS-1)*NB_FEATURES); - OPUS_COPY(&st->cont_features[(CONT_VECTORS-1)*NB_FEATURES], st->features, NB_FEATURES); + queue_features(st, st->features); get_fec_or_pred(st, st->features); - OPUS_MOVE(&st->cont_features[0], &st->cont_features[NB_FEATURES], (CONT_VECTORS-1)*NB_FEATURES); - OPUS_COPY(&st->cont_features[(CONT_VECTORS-1)*NB_FEATURES], st->features, NB_FEATURES); - fargan_cont(&st->fargan, st->pcm, st->cont_features); + queue_features(st, st->features); + fargan_cont(&st->fargan, &st->pcm[PLC_BUF_SIZE-FARGAN_CONT_SAMPLES], st->cont_features); } OPUS_MOVE(&st->plc_copy[1], &st->plc_copy[0], FEATURES_DELAY); st->plc_copy[0] = st->plc_net; @@ -217,18 +206,11 @@ int lpcnet_plc_conceal(LPCNetPLCState *st, opus_int16 *pcm) { if (st->loss_count >= 10) st->features[0] = MAX16(-10, st->features[0]+att_table[9] - 2*(st->loss_count-9)); else st->features[0] = MAX16(-10, st->features[0]+att_table[st->loss_count]); fargan_synthesize_int(&st->fargan, pcm, &st->features[0]); - { - float x[FRAME_SIZE]; - /* FIXME: Can we do better? */ - for (i=0;i<FRAME_SIZE;i++) x[i] = pcm[i]; - preemphasis(x, &st->enc.mem_preemph, x, PREEMPHASIS, FRAME_SIZE); - compute_frame_features(&st->enc, x); - process_single_frame(&st->enc, NULL); - } + lpcnet_compute_single_frame_features(&st->enc, pcm, st->features); OPUS_MOVE(&st->cont_features[0], &st->cont_features[NB_FEATURES], (CONT_VECTORS-1)*NB_FEATURES); - OPUS_COPY(&st->cont_features[(CONT_VECTORS-1)*NB_FEATURES], st->enc.features, NB_FEATURES); - OPUS_MOVE(st->pcm, &st->pcm[FRAME_SIZE], FARGAN_CONT_SAMPLES-FRAME_SIZE); - for (i=0;i<FRAME_SIZE;i++) st->pcm[FARGAN_CONT_SAMPLES-FRAME_SIZE+i] = (1.f/32768.f)*pcm[i]; + OPUS_COPY(&st->cont_features[(CONT_VECTORS-1)*NB_FEATURES], st->features, NB_FEATURES); + OPUS_MOVE(st->pcm, &st->pcm[FRAME_SIZE], PLC_BUF_SIZE-FRAME_SIZE); + for (i=0;i<FRAME_SIZE;i++) st->pcm[PLC_BUF_SIZE-FRAME_SIZE+i] = (1.f/32768.f)*pcm[i]; st->blend = 1; return 0; } diff --git a/dnn/lpcnet_private.h b/dnn/lpcnet_private.h index 15a74fefd89a7d10493b41283eb664ebdcbcf8dd..6302cc640bb4381d28ccd3557afc32192856b0f3 100644 --- a/dnn/lpcnet_private.h +++ b/dnn/lpcnet_private.h @@ -46,7 +46,7 @@ struct LPCNetEncState{ float burg_cepstrum[2*NB_BANDS]; }; -#define PLC_BUF_SIZE (FEATURES_DELAY*FRAME_SIZE + FRAME_SIZE) +#define PLC_BUF_SIZE (CONT_VECTORS*FRAME_SIZE) struct LPCNetPLCState { PLCModel model; FARGANState fargan; @@ -60,7 +60,7 @@ struct LPCNetPLCState { int fec_read_pos; int fec_fill_pos; int fec_skip; - float pcm[FARGAN_CONT_SAMPLES]; + float pcm[PLC_BUF_SIZE]; int blend; float features[NB_TOTAL_FEATURES]; float cont_features[CONT_VECTORS*NB_FEATURES]; diff --git a/src/opus_decoder.c b/src/opus_decoder.c index 0a500398c5c49348420f8c3757f3131ae79ac02c..1bdb6ad4ef882c3acb00e80d2de032e4719f081b 100644 --- a/src/opus_decoder.c +++ b/src/opus_decoder.c @@ -678,6 +678,7 @@ int opus_decode_native(OpusDecoder *st, const unsigned char *data, init_frames = (st->lpcnet.blend == 0) ? 2 : 0; features_per_frame = IMAX(1, frame_size/F10); needed_feature_frames = init_frames + features_per_frame; + lpcnet_plc_fec_clear(&st->lpcnet); for (i=0;i<needed_feature_frames;i++) { int feature_offset; /* We floor instead of rounding because 5-ms overlap compensates for the missing 0.5 rounding offset. */