audio_utils.py

import librosa
from IPython.display import Audio, display
import matplotlib.pyplot as plt
import torch
SAMPLE_RIR_URL = "https://pytorch-tutorial-assets.s3.amazonaws.com/VOiCES_devkit/distant-16k/room-response/rm1/impulse/Lab41-SRI-VOiCES-rm1-impulse-mc01-stu-clo.wav"

SAMPLE_NOISE_URL = "https://pytorch-tutorial-assets.s3.amazonaws.com/VOiCES_devkit/distant-16k/distractors/rm1/babb/Lab41-SRI-VOiCES-rm1-babb-mc01-stu-clo.wav"

def plot_spectrogram(spec, title=None, ylabel="freq_bin", aspect="auto", xmax=None):
    spec = spec.squeeze(0)
    spec = spec.numpy()
    fig, axs = plt.subplots(1, 1)
    axs.set_title(title or "Spectrogram (db)")
    axs.set_ylabel(ylabel)
    axs.set_xlabel("frame")
    im = axs.imshow(librosa.power_to_db(spec), origin="lower", aspect=aspect)
    if xmax:
        axs.set_xlim((0, xmax))
    fig.colorbar(im, ax=axs)
    plt.show(block=False)

def play_audio(waveform, sample_rate):
    waveform = waveform.numpy()

    num_channels, num_frames = waveform.shape
    if num_channels == 1:
        display(Audio(waveform[0], rate=sample_rate))
    elif num_channels == 2:
        display(Audio((waveform[0], waveform[1]), rate=sample_rate))
    else:
        raise ValueError("Waveform with more than 2 channels are not supported.")

def get_rir_sample(path, resample=None, processed=False):
    rir_raw, sample_rate = torch.load(path)
    if not processed:
        return rir_raw, sample_rate
    rir = rir_raw[:, int(sample_rate*1.01):int(sample_rate*1.3)]
    rir = rir / torch.norm(rir, p=2)
    rir = torch.flip(rir, [1])
    return rir, sample_rate