eden-emu/opus
15
0
Fork 0
mirror of https://github.com/xiph/opus.git synced 2025-05-31 23:57:42 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 4

Remove the need for useless exc and pred files

Browse source
This commit is contained in:
Jean-Marc Valin 2018-12-01 12:05:23 -05:00
parent b05f950e38
commit 91d90676e1
3 changed files with 11 additions and 21 deletions
Download patch file Download diff file Expand all files Collapse all files

4
dnn/README.md
View file

@ -19,7 +19,7 @@ This software is also a useful resource as an open source starting point for Wav
1. Then, run the resulting executable: 1. Then, run the resulting executable:
``` ```
./dump_data input.s16 exc.s8 features.f32 pred.s16 pcm.s16 ./dump_data input.s16 features.f32 pcm.s16
``` ```
where the first file contains 16 kHz 16-bit raw PCM audio (no header) where the first file contains 16 kHz 16-bit raw PCM audio (no header)
@ -29,7 +29,7 @@ always use ±5% or 10% resampling to augment your data).
1. Now that you have your files, you can do the training with: 1. Now that you have your files, you can do the training with:
``` ```
./train_lpcnet.py exc.s8 features.f32 pred.s16 pcm.s16 ./train_lpcnet.py features.f32 pcm.s16
``` ```
and it will generate a wavenet*.h5 file for each iteration. If it stops with a and it will generate a wavenet*.h5 file for each iteration. If it stops with a
"Failed to allocate RNN reserve space" message try reducing the *batch\_size* variable in train_wavenet_audio.py. "Failed to allocate RNN reserve space" message try reducing the *batch\_size* variable in train_wavenet_audio.py.

19
dnn/denoise.c
View file

@ -579,24 +579,20 @@ int main(int argc, char **argv) {
float mem_preemph=0; float mem_preemph=0;
float x[FRAME_SIZE]; float x[FRAME_SIZE];
FILE *f1; FILE *f1;
FILE *fexc;
FILE *ffeat; FILE *ffeat;
FILE *fpred;
FILE *fpcm; FILE *fpcm;
signed char iexc[FRAME_SIZE]; signed char iexc[FRAME_SIZE];
short pred[FRAME_SIZE]; short pred[FRAME_SIZE];
short pcm[FRAME_SIZE]; short pcm[FRAME_SIZE];
DenoiseState *st; DenoiseState *st;
st = rnnoise_create(); st = rnnoise_create();
if (argc!=6) { if (argc!=4) {
fprintf(stderr, "usage: %s <speech> <exc out> <features out> <prediction out> <pcm out> \n", argv[0]); fprintf(stderr, "usage: %s <speech> <features out>\n", argv[0]);
return 1; return 1;
} }
f1 = fopen(argv[1], "r"); f1 = fopen(argv[1], "r");
fexc = fopen(argv[2], "w"); ffeat = fopen(argv[2], "w");
ffeat = fopen(argv[3], "w"); fpcm = fopen(argv[3], "w");
fpred = fopen(argv[4], "w");
fpcm = fopen(argv[5], "w");
while (1) { while (1) {
kiss_fft_cpx X[FREQ_SIZE], P[WINDOW_SIZE]; kiss_fft_cpx X[FREQ_SIZE], P[WINDOW_SIZE];
float Ex[NB_BANDS], Ep[NB_BANDS]; float Ex[NB_BANDS], Ep[NB_BANDS];
@ -617,17 +613,14 @@ int main(int argc, char **argv) {
preemphasis(x, &mem_preemph, x, PREEMPHASIS, FRAME_SIZE); preemphasis(x, &mem_preemph, x, PREEMPHASIS, FRAME_SIZE);
compute_frame_features(st, iexc, pred, pcm, X, P, Ex, Ep, Exp, features, x); compute_frame_features(st, iexc, pred, pcm, X, P, Ex, Ep, Exp, features, x);
#if 1
fwrite(iexc, sizeof(signed char), FRAME_SIZE, fexc);
fwrite(features, sizeof(float), NB_FEATURES, ffeat); fwrite(features, sizeof(float), NB_FEATURES, ffeat);
fwrite(pred, sizeof(short), FRAME_SIZE, fpred);
fwrite(pcm, sizeof(short), FRAME_SIZE, fpcm); fwrite(pcm, sizeof(short), FRAME_SIZE, fpcm);
#endif
count++; count++;
} }
//fprintf(stderr, "matrix size: %d x %d\n", count, NB_FEATURES + 2*NB_BANDS + 1); //fprintf(stderr, "matrix size: %d x %d\n", count, NB_FEATURES + 2*NB_BANDS + 1);
fclose(f1); fclose(f1);
fclose(fexc); fclose(ffeat);
fclose(fpcm);
return 0; return 0;
} }

9
dnn/train_lpcnet.py
View file

@ -56,10 +56,8 @@ model, _, _ = lpcnet.new_lpcnet_model()
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['sparse_categorical_accuracy']) model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['sparse_categorical_accuracy'])
model.summary() model.summary()
exc_file = sys.argv[1] # not used at present feature_file = sys.argv[1]
feature_file = sys.argv[2] pcm_file = sys.argv[2] # 16 bit unsigned short PCM samples
pred_file = sys.argv[3] # LPC predictor samples. Not used at present, see below
pcm_file = sys.argv[4] # 16 bit unsigned short PCM samples
frame_size = 160 frame_size = 160
nb_features = 55 nb_features = 55
nb_used_features = model.nb_used_features nb_used_features = model.nb_used_features
@ -96,8 +94,7 @@ features = np.reshape(features, (nb_frames*feature_chunk_size, nb_features))
# Note: the LPC predictor output is now calculated by the loop below, this code was # Note: the LPC predictor output is now calculated by the loop below, this code was
# for an ealier version that implemented the prediction filter in C # for an ealier version that implemented the prediction filter in C
upred = np.fromfile(pred_file, dtype='int16') upred = np.zeros((nb_frames*pcm_chunk_size,), dtype='int16')
upred = upred[:nb_frames*pcm_chunk_size]
# Use 16th order LPC to generate LPC prediction output upred[] and (in # Use 16th order LPC to generate LPC prediction output upred[] and (in
# mu-law form) pred[] # mu-law form) pred[]