Spaces:
Build error
Build error
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,109 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import os
|
3 |
+
import subprocess
|
4 |
+
import tempfile
|
5 |
+
import matlab.engine
|
6 |
+
from st_audiorec import st_audiorec
|
7 |
+
import os.path
|
8 |
+
import numpy as np
|
9 |
+
import sounddevice as sd
|
10 |
+
from scipy.io.wavfile import write
|
11 |
+
from scipy.io import wavfile
|
12 |
+
|
13 |
+
|
14 |
+
|
15 |
+
def record(duration):
|
16 |
+
fs = 16000
|
17 |
+
seconds = duration
|
18 |
+
myrecording = sd.rec(int(seconds * fs), samplerate=fs, channels=1)
|
19 |
+
sd.wait()
|
20 |
+
write('clean_waveform.wav', fs, myrecording)
|
21 |
+
|
22 |
+
|
23 |
+
def main():
|
24 |
+
st.title("Upload WAV File, make the file noisy, run enhancement and transcribe")
|
25 |
+
uploaded_wav_file = st.file_uploader("Upload a WAV file", type=["wav"])
|
26 |
+
uploaded_noise_file = st.file_uploader("Upload a noise file", type = ["wav"])
|
27 |
+
snr = st.text_input("Enter SNR", "")
|
28 |
+
|
29 |
+
temp_dir = tempfile.mkdtemp()
|
30 |
+
if st.button("Record"):
|
31 |
+
record(5)
|
32 |
+
if st.button("Add Noise"):
|
33 |
+
wav_file_path = os.path.join(temp_dir, uploaded_wav_file.name)
|
34 |
+
with open(wav_file_path, "wb") as f1:
|
35 |
+
f1.write(uploaded_wav_file.getvalue())
|
36 |
+
|
37 |
+
noise_file_path = os.path.join(temp_dir, uploaded_noise_file.name)
|
38 |
+
with open(noise_file_path, "wb") as f2:
|
39 |
+
f2.write(uploaded_noise_file.getvalue())
|
40 |
+
|
41 |
+
#run_matlab_script(snr)
|
42 |
+
samplerate, signal = wavfile.read(wav_file_path)
|
43 |
+
samplerate, noise = wavfile.read(noise_file_path)
|
44 |
+
mix_audio(signal, noise, snr)
|
45 |
+
|
46 |
+
if st.button("Enhance"):
|
47 |
+
run_batch_script()
|
48 |
+
|
49 |
+
if st.button("Transcribe_zeroshot"):
|
50 |
+
transcribe_zeroshot()
|
51 |
+
if st.button("Transcribe_trained"):
|
52 |
+
transcribe_trained()
|
53 |
+
|
54 |
+
|
55 |
+
def run_matlab_script(snr):
|
56 |
+
|
57 |
+
read_fd, write_fd = os.pipe()
|
58 |
+
|
59 |
+
matlab_executable = 'matlab'
|
60 |
+
|
61 |
+
# Path to your MATLAB script
|
62 |
+
matlab_script = 'mixFiles.m'
|
63 |
+
|
64 |
+
# Data to send to MATLAB (replace with your actual data)
|
65 |
+
data_to_send = snr
|
66 |
+
|
67 |
+
# Run the MATLAB script
|
68 |
+
process = subprocess.Popen([matlab_executable, '-nodesktop', '-nosplash', '-r', f'run("{matlab_script}");exit;'], stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
|
69 |
+
|
70 |
+
# Send data to MATLAB
|
71 |
+
process.stdin.write(data_to_send)
|
72 |
+
process.stdin.close()
|
73 |
+
|
74 |
+
|
75 |
+
|
76 |
+
def run_batch_script():
|
77 |
+
command = r"OmniClear_cloud_demo noisy_waveform.wav enhanced_waveform.wav"
|
78 |
+
subprocess.run(command, shell=True)
|
79 |
+
|
80 |
+
def transcribe_zeroshot():
|
81 |
+
command = r"streamlit run loadlocal_zeroshot.py"
|
82 |
+
subprocess.run(command)
|
83 |
+
|
84 |
+
def transcribe_trained():
|
85 |
+
command = r"streamlit run loadlocal_trained.py"
|
86 |
+
subprocess.run(command)
|
87 |
+
|
88 |
+
|
89 |
+
def mix_audio(signal, noise, snr):
|
90 |
+
noise = noise[np.arange(len(signal)) % len(noise)]
|
91 |
+
noise = noise.astype(np.float32)
|
92 |
+
signal = signal.astype(np.float32)
|
93 |
+
signal_energy = np.mean(signal**2)
|
94 |
+
noise_energy = np.mean(noise**2)
|
95 |
+
snr = float(snr)
|
96 |
+
g = np.sqrt(10.0 ** (-snr/10) * signal_energy / noise_energy)
|
97 |
+
a = np.sqrt(1 / (1 + g**2))
|
98 |
+
b = np.sqrt(g**2 / (1 + g**2))
|
99 |
+
# mix the signals
|
100 |
+
rate = 16000
|
101 |
+
noisy_signal = a * signal + b * noise
|
102 |
+
scaled = np.int16(noisy_signal / np.max(np.abs(noisy_signal)) * 32767)
|
103 |
+
# Write the array to a WAV file
|
104 |
+
#print(scaled.shape)
|
105 |
+
write('noisy_waveform.wav', rate, scaled)
|
106 |
+
|
107 |
+
|
108 |
+
if __name__ == "__main__":
|
109 |
+
main()
|