def test_pitchgram_on_single_tone_should_have_peak_at_that_tone():
pitch = 12 + 7 # G5
f = Tuning().pitch_to_freq(pitch)
fs = 44100
x = sine(sample_time(0, 1, fs=fs), freq=f)
frame_size = 4096
hop_size = 2048
output_frame_size = hop_size
signal_frames = SignalFrames(x,
frame_size,
hop_size,
sample_rate=fs,
mono_mix=True)
bin_range = [-48, 67]
x_pitchgram = pitchgram(signal_frames,
output_frame_size,
magnitudes='power_db',
bin_range=bin_range,
bin_division=1)
max_bin_expected = pitch - bin_range[0]
max_bin_actual = x_pitchgram.mean(axis=0).argmax()
assert x_pitchgram.shape == (21, 115), x_pitchgram.shape
assert max_bin_actual == max_bin_expected
def test_reassigned_pitchgram_values_should_be_in_proper_range():
frame_size = 4096
hop_size = frame_size
output_frame_size = 1024
audio_file = os.path.join(DATA_DIR, 'she_brings_to_me.wav')
signal_frames = SignalFrames(audio_file, frame_size, hop_size, mono_mix=True)
X_r = pitchgram(signal_frames, output_frame_size, magnitudes='power_db')
assert np.all(X_r >= -120), 'min value: %f should be >= -120' % X_r.min()
assert np.all(X_r <= 0), 'max value: %f should be <= 0' % X_r.max()
def test_split_to_frames():
signal_frames = SignalFrames(np.arange(23),
frame_size=8,
hop_size=6,
sample_rate=44100)
assert np.allclose(
np.array([
[0, 1, 2, 3, 4, 5, 6, 7],
[6, 7, 8, 9, 10, 11, 12, 13],
[12, 13, 14, 15, 16, 17, 18, 19],
[18, 19, 20, 21, 22, 0, 0, 0],
]), signal_frames.frames)
def test_spectrogram_db_magnituds_should_be_in_proper_range():
frame_size = 4096
hop_size = 4096
audio_file = os.path.join(DATA_DIR, 'she_brings_to_me.wav')
signal_frames = SignalFrames(audio_file,
frame_size,
hop_size,
mono_mix=True)
w = create_window(frame_size)
X = stft_spectrogram(signal_frames.frames, w, magnitudes='power_db')
assert np.all(X >= -120), 'min value: %f should be >= -120' % X.min()
assert np.all(X <= 0), 'max value: %f should be <= 0' % X.max()