danstowell@0: 
danstowell@0: # utility functions
danstowell@0: 
danstowell@0: import numpy as np
danstowell@0: from numpy import float32
danstowell@0: 
danstowell@0: import os, errno
danstowell@0: from scikits.audiolab import Sndfile
danstowell@0: from scikits.audiolab import Format
danstowell@0: 
danstowell@0: from matplotlib.mlab import specgram
danstowell@0: 
danstowell@0: from userconfig import *
danstowell@0: 
danstowell@0: ########################################################
danstowell@0: 
danstowell@0: def standard_specgram(signal):
danstowell@0: 	"Return specgram matrix, made using the audio-layer config"
danstowell@0: 	return np.array(specgram(signal, NFFT=audioframe_len, noverlap=audioframe_len-audioframe_stride, window=np.hamming(audioframe_len))[0][specbinlow:specbinlow+specbinnum,:], dtype=float32)
danstowell@0: 
danstowell@0: def load_soundfile(inwavpath, startpossecs, maxdursecs=None):
danstowell@0: 	"""Loads audio data, optionally limiting to a specified start position and duration.
danstowell@0: 	Must be SINGLE-CHANNEL and matching our desired sample-rate."""
danstowell@0: 	framelen = 4096
danstowell@0: 	hopspls = framelen
danstowell@0: 	unhopspls = framelen - hopspls
danstowell@0: 	if (framelen % wavdownsample) != 0: raise ValueError("framelen needs to be a multiple of wavdownsample: %i, %i" % (framelen, wavdownsample))
danstowell@0: 	if (hopspls  % wavdownsample) != 0: raise ValueError("hopspls  needs to be a multiple of wavdownsample: %i, %i" % (hopspls , wavdownsample))
danstowell@0: 	if maxdursecs==None:
danstowell@0: 		maxdursecs = 9999
danstowell@0: 	sf = Sndfile(inwavpath, "r")
danstowell@0: 	splsread = 0
danstowell@0: 	framesread = 0
danstowell@0: 	if sf.channels != 1:       raise ValueError("Sound file %s has multiple channels (%i) - mono required." % (inwavpath, sf.channels))
danstowell@0: 	timemax_spls   = int(maxdursecs * sf.samplerate)
danstowell@0: 	if sf.samplerate != (srate * wavdownsample):
danstowell@0: 		raise ValueError("Sample rate mismatch: we expect %g, file has %g" % (srate, sf.samplerate))
danstowell@0: 	if startpossecs > 0:
danstowell@0: 			sf.seek(startpossecs * sf.samplerate) # note: returns IOError if beyond the end
danstowell@0: 	audiodata = np.array([], dtype=np.float32)
danstowell@0: 	while(True):
danstowell@0: 		try:
danstowell@0: 			if splsread==0:
danstowell@0: 				chunk = sf.read_frames(framelen)[::wavdownsample]
danstowell@0: 				splsread += framelen
danstowell@0: 			else:
danstowell@0: 				chunk = np.hstack((chunk[:unhopspls], sf.read_frames(hopspls)[::wavdownsample] ))
danstowell@0: 				splsread += hopspls
danstowell@0: 			framesread += 1
danstowell@0: 			if framesread % 25000 == 0:
danstowell@0: 				print("Read %i frames" % framesread)
danstowell@0: 			if len(chunk) != (framelen / wavdownsample):
danstowell@0: 				print("Not read sufficient samples - returning")
danstowell@0: 				break
danstowell@0: 			chunk = np.array(chunk, dtype=np.float32)
danstowell@0: 			audiodata = np.hstack((audiodata, chunk))
danstowell@0: 			if splsread >= timemax_spls:
danstowell@0: 				break
danstowell@0: 		except RuntimeError:
danstowell@0: 			break
danstowell@0: 	sf.close()
danstowell@0: 	return audiodata
danstowell@0: 
danstowell@0: def mkdir_p(path):
danstowell@0:     try:
danstowell@0:         os.makedirs(path)
danstowell@0:     except OSError as exc: # Python >2.5
danstowell@0:         if exc.errno == errno.EEXIST and os.path.isdir(path):
danstowell@0:             pass
danstowell@0:         else: raise
danstowell@0: