def one_song_fft(pcm1, song_name, subset=None):
# fig, axs = plt.subplots(nrows=2, ncols=1, figsize=(10, 4))
fig, axs = plt.subplots(2, 1, gridspec_kw={'height_ratios': [1, 2]}, figsize=(10, 6))
pcm1 = pcm.into_mono(pcm1)[0:len(pcm1) / 2]
if subset:
pcm1 = pcm1[subset[0]: subset[1]]
x = [i / 44100.0 for i in range(0, len(pcm1))]
y = pcm1
axs[0].plot(x, y, color='#006600', lw=0.05, antialiased=True)
axs[0].set_ylabel("Amplitude")
axs[0].set_xlabel("Time (Seconds)")
fft = np.fft.fft(pcm1)
fft = fft[0:len(fft)/2]
x = [pcm.frequency_at_fft_index(i, len(fft)) for i in range(0, len(fft))]
axs[1].set_ylim([-30000, 30000])
axs[1].set_xlim([0, 22000])
axs[1].plot(x, fft, color='#cc0000', lw=0.1, antialiased=True)
axs[1].set_ylabel("Amplitude")
axs[1].set_xlabel("Frequency (Hz)")
fig.tight_layout()
if subset:
file_path = 'graphs/one_song_fft_' + song_name + '_subset' + '.png'
else:
file_path = 'graphs/one_song_fft_' + song_name + '.png'
print file_path
plt.savefig(file_path)
def calculate(self, pcm_data):
# calculates the time domain zero crossings
pcm_mono = pcm.into_mono(pcm_data)
maximum = 0
summ = 0
for sample in pcm_mono:
summ += abs(sample)
maximum = max(maximum, abs(sample))
return (summ / float(maximum)) / len(pcm_mono)
def calculate(self, pcm_data):
sample_pack_size = 1024
pcm_mono = pcm.into_mono(pcm_data)
pcm_sample_packs = pcm.to_sample_packs(pcm_mono, sample_pack_size)
rolloffs = []
for sample_pack in pcm_sample_packs:
rolloffs.append(pcm.roll_off(sample_pack))
rolloff_sd = stats.standard_deviation(rolloffs)
return rolloff_sd
def calculate(self, pcm_data):
sample_pack_size = 1024
pcm_mono = pcm.into_mono(pcm_data)
pcm_sample_packs = pcm.to_sample_packs(pcm_mono, sample_pack_size)
centroids = []
for sample_pack in pcm_sample_packs:
centroids.append(pcm.centroid(sample_pack))
centroid_sd = stats.standard_deviation(centroids)
return centroid_sd
def calculate(self, pcm_data):
sample_pack_size = 1024
pcm_mono = pcm.into_mono(pcm_data)
pcm_sample_packs = pcm.to_sample_packs(pcm_mono, sample_pack_size)
centroids = []
for sample_pack in pcm_sample_packs:
centroids.append(pcm.centroid(sample_pack))
centroid_avg = stats.average(centroids)
return centroid_avg
def calculate(self, pcm_data):
# calculates the time domain zero crossings
pcm_mono = pcm.into_mono(pcm_data)
noise = 0
for i in range(1, len(pcm_mono)):
sign1 = 1 if pcm_mono[i] > 0 else 0
sign0 = 1 if pcm_mono[i - 1] > 0 else 0
noise += abs(sign1 - sign0)
return noise
def calculate(self, pcm_data):
sample_pack_size = 1024
pcm_mono = pcm.into_mono(pcm_data)
pcm_sample_packs = pcm.to_sample_packs(pcm_mono, sample_pack_size)
fluxes = []
for i in range(1, len(pcm_sample_packs)):
fluxes.append(pcm.spectral_flux(pcm_sample_packs[i - 1], pcm_sample_packs[i]))
fluxes_sd = stats.standard_deviation(fluxes)
return fluxes_sd
def one_song_fft_sub(pcm1, song_name, subset):
fig, axs = plt.subplots(2, 1, gridspec_kw={'height_ratios': [1, 2]}, figsize=(10, 6))
pcm1 = pcm.into_mono(pcm1)[0:len(pcm1) / 2]
pcm1 = pcm1[subset[0]: subset[1]]
x = [i / 44100.0 for i in range(0, len(pcm1))]
y = pcm1
# insert zeros
i = 0
while i < len(x):
x.insert(i, 0.0)
i += 2
axs[0].plot(x, y, color='#006600', lw=1)
# axs[0].bar(x, y, color='#006600', width=0.00001, linewidth=0)
axs[0].set_ylabel("Amplitude")
axs[0].set_xlabel("Time (Seconds)")
fft = np.fft.fft(pcm1)
fft = fft[0:len(fft)/2]
x = range(0, len(fft))
# axs[1].set_ylim([-30000, 30000])
# axs[1].set_xlim([0, 22000])
axs[1].bar(x, fft, color='#cc0000', antialiased=True)
axs[1].set_ylabel("Amplitude")
axs[1].set_xlabel("Frequency Index")
fig.tight_layout()
file_path = 'graphs/one_song_fft_' + song_name + '_subset' + '.png'
print file_path
plt.savefig(file_path)
def calculate(self, pcm_data):
pcm_mono = pcm.into_mono(pcm_data)
return bpm.determine_bpm(pcm_mono)
