def remove_noise(self):
y,sr = librosa.load("test.wav")
noise_len = 2 # seconds
noise = band_limited_noise(min_freq=2000, max_freq = 12000, samples=len(y), samplerate=sr)*10
noise_clip = noise[:sr*noise_len]
noise_reduced = nr.reduce_noise(audio_clip=y, noise_clip=noise_clip, prop_decrease=1.0, verbose=False)
sf.write('test.wav', noise_reduced, sr)
self.predict_lbl['text'] = 'Đã remove noise'
def add_noise(audio_clip, rate):
noise_len = 3
noise = band_limited_noise(
4000,
12000,
len(audio_clip),
rate
) * 10
noise_clip = noise[:rate * noise_len]
return audio_clip + noise, noise_clip
def test_reduce_generated_noise_stationary_without_noise_clip():
# load data
wav_loc = "assets/fish.wav"
rate, data = wavfile.read(wav_loc)
# add noise
noise_len = 2 # seconds
noise = (band_limited_noise(
min_freq=2000, max_freq=12000, samples=len(data), samplerate=rate) *
10)
audio_clip_band_limited = data + noise
return nr.reduce_noise(y=audio_clip_band_limited, sr=rate, stationary=True)
def test_reduce_generated_noise():
# load data
wav_loc = "assets/fish.wav"
rate, data = wavfile.read(wav_loc)
data = data / 32768.
# add noise
noise_len = 2 # seconds
noise = band_limited_noise(
min_freq=2000, max_freq=12000, samples=len(data), samplerate=rate) * 10
noise_clip = noise[:rate * noise_len]
audio_clip_band_limited = data + noise
return nr.reduce_noise(audio_clip=audio_clip_band_limited,
noise_clip=noise_clip,
verbose=True)
def test_reduce_generated_noise():
# load data
wav_loc = "assets/coffe-1_2020-03-08_170854361.wav"
rate, data = wavfile.read(wav_loc)
data = int16_to_float32(data)
# add noise
noise_len = 2 # seconds
noise = (band_limited_noise(
min_freq=2000, max_freq=12000, samples=len(data), samplerate=rate) *
10)
noise_clip = noise[:rate * noise_len]
audio_clip_band_limited = data + noise
return nr.reduce_noise(audio_clip=audio_clip_band_limited,
noise_clip=noise_clip,
verbose=True)
def test_reduce_generated_noise_batches():
# load data
wav_loc = "assets/fish.wav"
rate, data = wavfile.read(wav_loc)
# add noise
noise_len = 2 # seconds
noise = (band_limited_noise(
min_freq=2000, max_freq=12000, samples=len(data), samplerate=rate) *
10)
noise_clip = noise[:rate * noise_len]
audio_clip_band_limited = data + noise
return nr.reduce_noise(y=audio_clip_band_limited,
sr=rate,
stationary=False,
chunk_size=30000)
import numpy as np
import io
#ucitavanje zvuka
wav_loc = "C:/Users/Djordje/Desktop/download.wav"
rate, data = wavfile.read(wav_loc)
data = data / 32768
IPython.display.Audio(data=data, rate=rate)
fig, ax = plt.subplots(figsize=(20, 3))
ax.plot(data)
#dodavanje suma
noise_len = 2 # sekunde
noise = band_limited_noise(
min_freq=2000, max_freq=12000, samples=len(data), samplerate=rate) * 10
noise_clip = noise[:rate * noise_len]
audio_clip_band_limited = data + noise
fig, ax = plt.subplots(figsize=(20, 3))
ax.plot(audio_clip_band_limited)
IPython.display.Audio(data=audio_clip_band_limited, rate=rate)
#uklanjanje suma
noise_reduced = nr.reduce_noise(audio_clip=audio_clip_band_limited,
noise_clip=noise_clip,
prop_decrease=1.0,
verbose=True)
#zvuk posle uklonjenog suma