import random
import numpy as np
import librosa
import scipy.signal
from sklearn.utils import gen_even_slices
def compute_stft(audio, window_size, hop_size, log=True, eps=1e-4):
f, t, s = scipy.signal.stft(
audio, nperseg=window_size, noverlap=hop_size)
s = np.abs(s)
if log:
s = np.log(s + eps)
return s
def trim_audio(audio):
audio, interval = librosa.effects.trim(audio, top_db=60)
return audio
def read_audio(file):
audio, sr = librosa.load(file, sr=None)
return audio, sr
def mix_audio_and_labels(first_audio, second_audio, first_labels, second_labels):
new_labels = np.clip(first_labels + second_labels, 0, 1)
a = np.random.uniform(0.4, 0.6)
shorter, longer = first_audio, second_audio
if shorter.size == longer.size:
return (shorter + longer) / 2, new_labels
if first_audio.size > second_audio.size:
shorter, longer = longer, shorter
start = random.randint(0, longer.size - 1 - shorter.size)
end = start + shorter.size
longer *= a
longer[start:end] =+ shorter * (1 - a)
return longer, new_labels
def shuffle_audio(audio, chunk_length=0.5, sr=None):
n_chunks = int((audio.size / sr) / chunk_length)
if n_chunks in (0, 1):
return audio
slices = list(gen_even_slices(audio.size, n_chunks))
random.shuffle(slices)
shuffled = np.concatenate([audio[s] for s in slices])
return shuffled
def cutout(audio, area=0.25):
area = int(audio.size * area)
start = random.randrange(audio.size)
end = start + area
audio[start:end] = 0
return audio
