Easily plot the magnitude spectrum of any audio signal and save it as a png file using Python and Matplotlib.Pyplot library. Includes styling the figure so that it looks reasonably well. You can use decibels full scale (dBFS) for the magnitude axis and a logarithmic frequency axis with ISO-standardized frequencies. Feel free to copy, paste & mod…

plot_magnitude_spectrum.py

import matplotlib.pyplot as plt
import numpy as np
from pathlib import Path
import soundfile as sf
import librosa


COLOR = "#ef7600"
IMG_OUTPUT_PATH = Path("img")
SAVE_PARAMS = {"dpi": 300, "bbox_inches": "tight", "transparent": True}

XTICKS = np.array([31.25, 62.5, 125, 250, 500, 1000, 2000, 4000, 8000])
XTICK_LABELS = np.array(["31.25", "62.5", "125", "250", "500", "1k", "2k", "4k", "8k"])

plt.rcParams.update({"font.size": 20})


def save_spectrum(output_path):
    output_path.parent.mkdir(parents=True, exist_ok=True)
    plt.savefig(output_path, **SAVE_PARAMS)


def plot_spectrum_and_save(frequencies, magnitude_spectrum, output_path: Path):
    plt.figure(figsize=(12, 6))
    plt.plot(frequencies, magnitude_spectrum, COLOR)
    xlim = [frequencies[0], frequencies[-1]]
    plt.xlim(xlim)
    plt.xlabel("frequency [Hz]")
    plt.hlines(0, xlim[0], xlim[1], colors="k")
    plt.xticks(None, None)
    plt.yticks([])
    plt.ylabel("magnitude")
    ax = plt.gca()
    ax.spines["top"].set_visible(False)
    ax.spines["right"].set_visible(False)
    ax.spines["bottom"].set_visible(False)

    save_spectrum(output_path)
    plt.close()


def plot_spectrum_db_and_save(frequencies, magnitude_spectrum, output_path: Path):
    plt.figure(figsize=(12, 6))
    plt.plot(frequencies, magnitude_spectrum, COLOR)
    plt.xlim([0, frequencies[-1]])
    plt.ylim([-60, 0])
    plt.grid()
    plt.xlabel("frequency [Hz]")
    plt.ylabel("magnitude [dBFS]")
    ax = plt.gca()
    ax.spines["top"].set_visible(False)
    ax.spines["right"].set_visible(False)

    save_spectrum(output_path)
    plt.close()


def plot_spectrum_db_in_octaves_and_save(
    frequencies, magnitude_spectrum, output_path: Path
):
    plt.figure(figsize=(12, 6))
    plt.semilogx(frequencies, magnitude_spectrum, COLOR)
    plt.ylim([-60, 0])
    plt.xticks(XTICKS, XTICK_LABELS)
    min_x = 29
    plt.xlim([min_x, frequencies[-1]])
    plt.grid()
    plt.xlabel("frequency [Hz]")
    plt.ylabel("magnitude [dBFS]")
    ax = plt.gca()
    ax.spines["top"].set_visible(False)
    ax.spines["right"].set_visible(False)

    # plt.show() # closes the figure
    save_spectrum(output_path)
    plt.close()


def main():
    signal, sample_rate = sf.read(
        Path(".") / "data" / "LibriSpeech-84-121123-0001.flac"
    )
    print(f"Signal's sample rate: {sample_rate} Hz.")
    magnitude_spectrum = np.abs(np.fft.rfft(signal))
    frequencies = np.fft.rfftfreq(signal.shape[0], 1 / sample_rate)
    plot_spectrum_and_save(
        frequencies,
        magnitude_spectrum,
        IMG_OUTPUT_PATH / "speech_magnitude_spectrum.png",
    )
    normalized_magnitude_spectrum = magnitude_spectrum / np.amax(magnitude_spectrum)
    magnitude_spectrum_db = librosa.amplitude_to_db(normalized_magnitude_spectrum)
    plot_spectrum_db_and_save(
        frequencies,
        magnitude_spectrum_db,
        IMG_OUTPUT_PATH / "speech_magnitude_spectrum_db.png",
    )
    plot_spectrum_db_in_octaves_and_save(
        frequencies,
        magnitude_spectrum_db,
        IMG_OUTPUT_PATH / "speech_magnitude_spectrum_db_in_octaves.png",
    )


if __name__ == "__main__":
    main()