A pip installable package for this code is now available at https://github.com/timsainb/noisereduce

Noise reduction in python using¶

This algorithm is based (but not completely reproducing) on the one outlined by Audacity for the noise reduction effect (Link to C++ code)
The algorithm requires two inputs:
1. A noise audio clip comtaining prototypical noise of the audio clip
2. A signal audio clip containing the signal and the noise intended to be removed

Steps of algorithm¶

An FFT is calculated over the noise audio clip
Statistics are calculated over FFT of the the noise (in frequency)
A threshold is calculated based upon the statistics of the noise (and the desired sensitivity of the algorithm)
An FFT is calculated over the signal
A mask is determined by comparing the signal FFT to the threshold
The mask is smoothed with a filter over frequency and time
The mask is appled to the FFT of the signal, and is inverted

In [1]:

import IPython
from scipy.io import wavfile
import scipy.signal
import numpy as np
import matplotlib.pyplot as plt
import librosa
%matplotlib inline

Load data¶

In [2]:

wav_loc = "assets/audio/fish.wav"
rate, data = wavfile.read(wav_loc)
data = data / 32768

In [3]:

# from https://stackoverflow.com/questions/33933842/how-to-generate-noise-in-frequency-range-with-numpy
def fftnoise(f):
    f = np.array(f, dtype="complex")
    Np = (len(f) - 1) // 2
    phases = np.random.rand(Np) * 2 * np.pi
    phases = np.cos(phases) + 1j * np.sin(phases)
    f[1 : Np + 1] *= phases
    f[-1 : -1 - Np : -1] = np.conj(f[1 : Np + 1])
    return np.fft.ifft(f).real


def band_limited_noise(min_freq, max_freq, samples=1024, samplerate=1):
    freqs = np.abs(np.fft.fftfreq(samples, 1 / samplerate))
    f = np.zeros(samples)
    f[np.logical_and(freqs >= min_freq, freqs <= max_freq)] = 1
    return fftnoise(f)

In [4]:

IPython.display.Audio(data=data, rate=rate)

Out[4]:

In [5]:

fig, ax = plt.subplots(figsize=(20,4))
ax.plot(data)

Out[5]:

[<matplotlib.lines.Line2D at 0x7f67290bac50>]

Add noise¶

In [6]:

noise_len = 2 # seconds
noise = band_limited_noise(min_freq=4000, max_freq = 12000, samples=len(data), samplerate=rate)*10
noise_clip = noise[:rate*noise_len]
audio_clip_band_limited = data+noise

In [7]:

fig, ax = plt.subplots(figsize=(20,4))
ax.plot(audio_clip_band_limited)
IPython.display.Audio(data=audio_clip_band_limited, rate=rate)

Out[7]: