Spaces:
Runtime error
Runtime error
File size: 3,015 Bytes
b87f798 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 |
"""Module containing audio helper functions.
"""
import numpy as np
import config as cfg
RANDOM = np.random.RandomState(cfg.RANDOM_SEED)
def openAudioFile(path: str, sample_rate=cfg.SAMPLE_RATE, offset=0.0, duration=None):
"""Open an audio file.
Opens an audio file with librosa and the given settings.
Args:
path: Path to the audio file.
sample_rate: The sample rate at which the file should be processed.
offset: The starting offset.
duration: Maximum duration of the loaded content.
Returns:
Returns the audio time series and the sampling rate.
"""
# Open file with librosa (uses ffmpeg or libav)
import librosa
sig, rate = librosa.load(path, sr=sample_rate, offset=offset, duration=duration, mono=True, res_type="kaiser_fast")
return sig, rate
def saveSignal(sig, fname: str):
"""Saves a signal to file.
Args:
sig: The signal to be saved.
fname: The file path.
"""
import soundfile as sf
sf.write(fname, sig, cfg.SAMPLE_RATE, "PCM_16")
def noise(sig, shape, amount=None):
"""Creates noise.
Creates a noise vector with the given shape.
Args:
sig: The original audio signal.
shape: Shape of the noise.
amount: The noise intensity.
Returns:
An numpy array of noise with the given shape.
"""
# Random noise intensity
if amount == None:
amount = RANDOM.uniform(0.1, 0.5)
# Create Gaussian noise
try:
noise = RANDOM.normal(min(sig) * amount, max(sig) * amount, shape)
except:
noise = np.zeros(shape)
return noise.astype("float32")
def splitSignal(sig, rate, seconds, overlap, minlen):
"""Split signal with overlap.
Args:
sig: The original signal to be split.
rate: The sampling rate.
seconds: The duration of a segment.
overlap: The overlapping seconds of segments.
minlen: Minimum length of a split.
Returns:
A list of splits.
"""
sig_splits = []
for i in range(0, len(sig), int((seconds - overlap) * rate)):
split = sig[i : i + int(seconds * rate)]
# End of signal?
if len(split) < int(minlen * rate):
break
# Signal chunk too short?
if len(split) < int(rate * seconds):
split = np.hstack((split, noise(split, (int(rate * seconds) - len(split)), 0.5)))
sig_splits.append(split)
return sig_splits
def cropCenter(sig, rate, seconds):
"""Crop signal to center.
Args:
sig: The original signal.
rate: The sampling rate.
seconds: The length of the signal.
"""
if len(sig) > int(seconds * rate):
start = int((len(sig) - int(seconds * rate)) / 2)
end = start + int(seconds * rate)
sig = sig[start:end]
# Pad with noise
elif len(sig) < int(seconds * rate):
sig = np.hstack((sig, noise(sig, (int(seconds * rate) - len(sig)), 0.5)))
return sig
|