def compute_ssm(wav_file, h, ssm_read_pk, is_ismir=False, tonnetz=False):
"""Computes the self similarity matrix from an audio file.
Parameters
----------
wav_file: str
Path to the wav file to be read.
h : float
Hop size.
ssm_read_pk : bool
Whether to read the ssm from a pickle file or not (note: this function
utomatically saves the ssm in a pickle file).
is_ismir : bool
Produce the plots that appear on the ISMIR paper.
tonnetz : bool
Compute tonnetz instead of Chroma features.
Returns
-------
X : np.array((N, N))
Self-similarity matrix
"""
if not ssm_read_pk:
# Read WAV file
logging.info("Reading the WAV file...")
C = utils.compute_audio_chromagram(wav_file, h)
C = utils.median_filter(C, L=9)
if is_ismir:
ismir.plot_chroma(C)
# Compute Tonnetz if needed
F = C
if tonnetz:
F = utils.chroma_to_tonnetz(C)
# Compute the self similarity matrix
logging.info("Computing key-invariant self-similarity matrix...")
X = utils.compute_key_inv_ssm(F, h)
#plt.imshow(X, interpolation="nearest", aspect="auto")
#plt.show()
utils.write_cPickle(wav_file + "-audio-ssm.pk", X)
else:
X = utils.read_cPickle(wav_file + "-audio-ssm.pk")
if is_ismir:
#X = X**2.5
ismir.plot_ssm(X)
ismir.plot_score_examples(X)
return X
def compute_beatsync_features(ticks, audio):
"""Computes the HPCP and MFCC beat-synchronous features given a set
of beats (ticks)."""
MFCC = STFTFeature(FRAME_SIZE, HOP_SIZE, WINDOW_TYPE,
ES.MFCC(numberCoefficients=14), ticks, SAMPLE_RATE)
HPCP = STFTFeature(FRAME_SIZE, HOP_SIZE, WINDOW_TYPE, ES.HPCP(),
ticks, SAMPLE_RATE)
logging.info("Computing Beat-synchronous MFCCs...")
mfcc = MFCC.compute_features(audio)
logging.info("Computing Beat-synchronous HPCPs...")
hpcp = HPCP.compute_features(audio)
logging.info("Computing Beat-synchronous Tonnetz...")
tonnetz = utils.chroma_to_tonnetz(hpcp)
return mfcc.tolist(), hpcp.tolist(), tonnetz.tolist()
