FreeDV
/
phasenn
mirror of https://github.com/drowe67/phasenn.git
1
0
Fork 0
Code Issues Releases Wiki Activity

quite reaonable results on visible inspection of a few frames

master
David 2019-12-13 17:36:32 +10:30
parent 76ffb675e4
commit d372e37903
1 changed files with 15 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
newamp1_train.py 100644 → 100755
View File

@ -26,7 +26,7 @@ os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
width = 256 width = 256
nb_batch = 32 nb_batch = 32
K = 20 newamp1_K = 20
nb_plots = 6 nb_plots = 6
def list_str(values): def list_str(values):
@ -37,6 +37,7 @@ parser.add_argument('featurefile', help='f32 file of newamp1 rate K vectors')
parser.add_argument('modelfile', help='Codec 2 model records with rate L vectors') parser.add_argument('modelfile', help='Codec 2 model records with rate L vectors')
parser.add_argument('--frames', type=list_str, default="30,31,32,33,34,35", help='Frames to view') parser.add_argument('--frames', type=list_str, default="30,31,32,33,34,35", help='Frames to view')
parser.add_argument('--epochs', type=int, default=10, help='Number of training epochs') parser.add_argument('--epochs', type=int, default=10, help='Number of training epochs')
parser.add_argument('--nb_samples', type=int, default=1000000, help='Number of frames to train on')
args = parser.parse_args() args = parser.parse_args()
assert nb_plots == len(args.frames) assert nb_plots == len(args.frames)
@ -47,26 +48,32 @@ nb_voiced = np.count_nonzero(voiced)
print("nb_samples: %d voiced %d" % (nb_samples, nb_voiced)) print("nb_samples: %d voiced %d" % (nb_samples, nb_voiced))
# read in rate K vectors # read in rate K vectors
features = np.fromfile(args.feature_file, dtype='float32') features = np.fromfile(args.featurefile, dtype='float32')
nb_features = 1+K+K nb_features = 1 + newamp1_K + newamp1_K
nb_samples1 = len(features)/nb_features nb_samples1 = len(features)/nb_features
print("nb_samples1: %d" % (nb_samples)) print("nb_samples1: %d" % (nb_samples))
assert nb_samples == nb_samples1 assert nb_samples == nb_samples1
features = np.reshape(features, (nb_samples, nb_features)) features = np.reshape(features, (nb_samples, nb_features))
print(features.shape) print(features.shape)
rateK = features[:,1:K+1] rateK = features[:,newamp1_K+1:]
print(rateK.shape)
# find ans subtract mean for each frame
mean_amp = np.zeros(nb_samples)
for i in range(nb_samples):
mean_amp[i] = np.mean(np.log10(A[i,1:L[i]+1]))
# set up sparse amp output vectors # set up sparse amp output vectors
amp_sparse = np.zeros((nb_samples, width)) amp_sparse = np.zeros((nb_samples, width))
for i in range(nb_samples): for i in range(nb_samples):
for m in range(1,L[i]+1): for m in range(1,L[i]+1):
bin = int(np.round(m*Wo[i]*width/np.pi)); bin = min(width-1, bin) bin = int(np.round(m*Wo[i]*width/np.pi)); bin = min(width-1, bin)
amp_sparse[i,bin] = 20*np.log10(A[i,m]) amp_sparse[i,bin] = np.log10(A[i,m]) - mean_amp[i]
# our model # our model
model = models.Sequential() model = models.Sequential()
model.add(layers.Dense(pairs, activation='relu', input_dim=K)) model.add(layers.Dense(4*newamp1_K, activation='relu', input_dim=newamp1_K))
model.add(layers.Dense(4*K, activation='relu')) model.add(layers.Dense(4*newamp1_K, activation='relu'))
model.add(layers.Dense(width)) model.add(layers.Dense(width))
model.summary() model.summary()
@ -119,7 +126,7 @@ plt.title('Amplitudes Spectra')
for r in range(nb_plots): for r in range(nb_plots):
plt.subplot(nb_plotsy,nb_plotsx,r+1) plt.subplot(nb_plotsy,nb_plotsx,r+1)
f = frames[r]; f = frames[r];
plt.plot(20*np.log10(A[f,1:L[f]]),'g') plt.plot(np.log10(A[f,1:L[f]])-mean_amp[f],'g')
plt.plot(amp_est[f,1:L[f]],'r') plt.plot(amp_est[f,1:L[f]],'r')
t = "frame %d" % (f) t = "frame %d" % (f)
plt.title(t) plt.title(t)