def to_wav(src_path, wav_path):
sox_in = pysox.CSoxStream(src_path)
sox_out = pysox.CSoxStream(wav_path, 'w', pysox.CSignalInfo(16000, 1, 16), fileType='wav')
sox_chain = pysox.CEffectsChain(sox_in, sox_out)
sox_chain.add_effect(pysox.CEffect('rate', ['16000']))
sox_chain.flow_effects()
sox_out.close()
def downsample(fs, sig):
in_file = random_string() + ".wav"
out_file = random_string() + ".wav"
frame_len = fs * WINDOW_SIZE
pad = len(sig) % frame_len
if pad > 0:
sig = np.append(sig, np.zeros(frame_len - pad))
f = Sndfile(in_file, 'w',
Format(type="wav", encoding='pcm16', endianness="file"), 1, fs)
f.write_frames(sig)
f.close()
sox_in = pysox.CSoxStream(in_file)
sox_out = pysox.CSoxStream(out_file,
'w',
pysox.CSignalInfo(SAMPLE_RATE, 1, 8),
fileType='wav')
sox_chain = pysox.CEffectsChain(sox_in, sox_out)
sox_chain.add_effect(pysox.CEffect("rate", [str(SAMPLE_RATE)]))
sox_chain.flow_effects()
sox_out.close()
f = Sndfile(out_file, 'r')
sig = f.read_frames(f.nframes)
f.close()
os.unlink(in_file)
os.unlink(out_file)
return sig
infile = pysox.CSoxStream(file)
outfile = pysox.CSoxStream(file + "-converted.flac", "w", infile.get_signal_info())
chain = pysox.CEffectsChain(infile, outfile)
chain.flow_effects()
outfile.close()
# effects chain
def fileEffect(filestoconvert)
for file in filestoconvert:
infile = pysox.CSoxStream(file)
# gather info on file
samplerate = infile.get_signal().get_signalinfo()['rate']
precision = infile.get_signal().get_signalinfo()['precision']
# fix sample rate and precision
if samplerate < minsamplerate and precision < minprecision:
outfile = pysox.CSoxStream(file + "-upsampled-higher_bit.flac", "w", pysox.CSignalInfo(minsamplerate, 2, minprecision))
# fix precision only
elif samplerate > minsamplerate and precision < minprecision:
outfile = pysox.CSoxStream(file + "-higher_bit.flac", "w", pysox.CSignalInfo(samplerate, 2, minprecision))
# fix sample rate only
elif samplerate < minsamplerate and precision > minprecision:
outfile = pysox.CSoxStream(file + "-upsampled.flac", "w", pysox.CSignalInfo(minsamplerate, 2, precision))
chain = pysox.CEffectsChain(infile, outfile)
chain.flow_effects()
outfile.close()
os.remove(file)
def main():
checkArgs()
# convert lossys
lossyfilestoconvert = fileList(lossys)
parser.add_argument('-m, --model-file',
action='store',
type=str,
dest='model_file',
help='the model file to use',
default='models/params.pkl')
args = parser.parse_args()
# downsample file to 8KHz, 8 bits per sample
in_file = args.input_file
out_file = random_string() + ".wav"
sox_in = pysox.CSoxStream(in_file)
sox_out = pysox.CSoxStream(out_file,
'w',
pysox.CSignalInfo(SAMPLE_RATE, 1, 8),
fileType='wav')
sox_chain = pysox.CEffectsChain(sox_in, sox_out)
sox_chain.add_effect(pysox.CEffect("rate", [str(SAMPLE_RATE)]))
sox_chain.flow_effects()
sox_out.close()
input_file = Sndfile(out_file, 'r')
fs = input_file.samplerate
num_frames = input_file.nframes
window = args.window_size / 1000.
chunk_size = int(np.floor(window * fs))
mlp = MLP_VAD(args.model_file)
