Python audiolab.wavread函数代码示例

def main():

    # import soundfile
    snd = wavread('trumpet.wav')[0]
    kick = wavread('kick.wav')[0]
    amb = wavread('amb.wav')[0]
    amb = amb * 0.8                 # reduce gain of this soundfile a little bit
    
    print len(amb)
    #low_demo(snd, 10., 500.)
    #high_demo(snd, 10000., 10.)
    #allpass_demo(snd, 1000, -find_c(1000., fs), find_c(1000., fs), 1.0)
    #iir_comb_demo(kick, 100, 0.5, -0.5)

    t = len(amb) / fs
    period = 1.0 / fs
    t_v = arange(0.0, t, period)
    delayTime = 2.0
    width = 1.0
    freq = 1
    breakPoint = (sin(2. * pi * freq * t_v))
    #breakPoint = linspace(1, -1, len(amb))

    #var_allpass_demo(snd, delayTime / 1000., width / 1000., -find_c(8000, fs), find_c(8000, fs), 1.0, breakPoint)
    #var_allpass_demo(amb, delayTime / 1000., width / 1000., 0.5, -0.5, 0.0, breakPoint)

    # flanger
    var_allpass_demo(amb, delayTime, width, 0.7, 0.7, 0.7, breakPoint)

def noise_reduce_test():
	sample = wavread('../../sounds/single-bloop-trimmed.wav')[0]
	noise = wavread('../../sounds/single-bloop-noise.wav')[0]
	sample = bandpass(sample,30000,50000)
	t0 = time.time()
	sample = noise_reduce(sample,noise,NoiseReduceSettings())
	print 'noise filter in time:', round(time.time() - t0,2)
	'''

def sound_wav():
    clf()
    (snd, sampFreq, nBits) = audiolab.wavread('temp.wav')
    wave_form = []
    signal = snd[:,0]
    if (len(signal)) < 500000:
        timeArray = arange(0, float(len(signal)), 1)
        timeArray = timeArray / sampFreq
        wave_form = signal
    else:
        downsample_factor = len(signal) / 30000
        i = 0
        while i < len(signal):
            wave_form = wave_form + [signal[i]]
            i = i + downsample_factor
        timeArray = arange(0, float(len(wave_form)), 1)
        timeArray = timeArray * downsample_factor / sampFreq
    timeArray = timeArray * 1000
    plot(timeArray, wave_form, color='k')
    ylabel('Amplitude')
    xlabel('Time (ms)')
    savefig('wave_form.png', bbox_inches=0)
    # show()
# setup('skream.wav')
# sound_wav()
# teardown()

 def __init__(self, filepath):
     
     self.filepath = filepath
     (self.audio_array, self.sample_rate, self.format) = wavread(filepath)
     self.name = os.path.basename(filepath)
     samples = len(self.audio_array)
     self.length = float(samples) / float(self.sample_rate)

def wavread(path):
    """
    Wrapper around scikits functions
    Returns: wavdata, sample rate, encoding type
    See pyaudiolab or scikits.audiolab for more information
    """
    return AUDIOLAB.wavread(path)

    def test_bad_wavread(self):
        """ Check wavread on bad file"""
        # Create a tmp audio file with non wav format, write some random data into it,
        # and check it can not be opened by wavread
        rfd, fd, cfilename   = open_tmp_file('pysndfiletest.wav')
        try:
            nbuff = 22050
            noise = 0.1 * N.random.randn(nbuff)

            # Open the copy file for writing
            format = audio_format('aiff', 'pcm16')
            b = Sndfile(cfilename, 'w', format, 1, nbuff)

            b.write_frames(noise)

            b.close()

            b = Sndfile(cfilename, 'r')
            rcnoise = b.read_frames(nbuff)
            b.close()

            try:
                rnoise  = wavread(cfilename)[0]
                raise Exception("wavread on non wav file succeded, expected to fail")
            except ValueError, e:
                pass
                #print str(e) + ", as expected"

        finally:
            close_tmp_file(rfd, cfilename)

def computeFeaturesForFullSong(file_path, feature_list, pack_size):
    """
    Computes each of the features (must be full_song features) for the song recording.
    This method is used for one shot computation of a songs features.
    :param file_path:
    :param features:
    :param pack_size:
    :return: a tuple of values with length = len(features). Each item is the resulting feature value corresponding to features[].
    """

    # will hold the evaluated feature values
    feature_values = []

    raw_data, fs, enc = wavread(file_path)
    raw_chunks = chunks(raw_data, pack_size)

    for feature_name in feature_list:
        # print "Computing " + feature_name
        class_ = getattr(features, feature_name)
        if class_.requireFullSong is False: # ensure full song
            raise "Every feature must be a full song feature"

        feature = class_(raw_chunks)
        feature_values.append(feature.value)

    return feature_values

def convert_wav(File, ofile):
    import scikits.audiolab as audiolab
    from scikits.samplerate import resample
    # lastest scikits.audiolab include sound record lib, based on python-alsaaudio
    # if you want make the down sample rate using scipy.signal
    #import scipy.signal

    #using audiolab to read wav file
    Signal, fs = audiolab.wavread(File)[:2]
    #changing the original sample rate to 16000fs fast mode
    Signal = resample(Signal, fr/float(fs), 'sinc_best')
     
    #changing sample rate from audio file using scipy this is a bit slow
    #Signal=scipy.signal.resample(Signal,int(round(len(Getsignal)*fr)/float(fs)),window=None)
     
    # file Format type
    fmt = audiolab.Format('flac', 'pcm16')
    nchannels   = 1
     
    # convert into the file .flac
    ofile =  audiolab.Sndfile(FileNameTmp, 'w', fmt, nchannels, fr)
     
    #writing in the file
    ofile.write_frames(Signal)
    #
    return ofile

def gather_training_data(path=SAMPLE_PATH):
    instr_names = os.walk(path).next()[1]
    samples = dict()

    pitch_pattern = re.compile("([A-G][sb]?)(\d+)")

    # NOTE: Could potentially make subdirs for different qualities

    for instr in instr_names:
        #if instr not in ('guitar', 'trumpet'): continue
        instr_samples = []
        instr_sample_dir = "%s\%s" % (SAMPLE_PATH, instr)
        for samp in [f for f in os.listdir(instr_sample_dir) \
                if os.path.isfile(os.path.join(instr_sample_dir, f)) \
                and os.path.splitext(f)[1].lower() == ".wav"]:
            data, fs, enc = skal.wavread("%s\%s" % (instr_sample_dir, samp))

            matches = pitch_pattern.search(samp)
            assert matches is not None

            chroma, octave = matches.groups()
            chroma = canonical_chroma[chroma]

            # NOTE: It's quite possible that using a dictionary
            #       instead of a list will be helpful, but we'll
            #       cross that bridge when we get to it
            instr_samples.append( (data, chroma, octave) )

        samples[instr] = instr_samples

    return samples

def main():
    """
    Main function for processing the specified soundfile through this reverb.
    """

    parser = argparse.ArgumentParser(description='Artificial Reverb')
    parser.add_argument('soundfile', help='audio file to process', type=validInput)        # the soundfile is the first agument, with parameter values to follow
    parser.add_argument('outfile', help='path to output file', type=validInput)
    parser.add_argument('-w', '--wetdry', default=0.2, type=float, help='amount of wet signal in the mix')
    parser.add_argument('-da', '--damping', default=0.25, type=float, help='amount of high frequency damping')
    parser.add_argument('-de', '--decay', default=0.4, type=float, help='amount of attentuation applied to signal to make it decay')
    parser.add_argument('-pd', '--predelay', default=30, type=float, help='amount of time before starting reverb')
    parser.add_argument('-b', '--bandwidth', default=0.6, type=float, help='amount of high frequency attentuation on input')
    parser.add_argument('-t', '--tankoffset', default=0, type=float, help='amount of time (ms) to increase the last tank delay time')

    # Parse the commandline arguments
    args = parser.parse_args()

    # Get the entire path and assign soundfile
    soundfilePath = os.path.join(os.getcwd(), args.soundfile)
    
    # From here on, x refers to the input signal
    x, sampleRate, wavType = wavread(soundfilePath)
    dry = x.copy()

    y = reverbTest(x, sampleRate, args.damping, args.decay, args.predelay, args.bandwidth, args.tankoffset)

    # Apply wet/dry mix
    output = dryWet(dry, y, args.wetdry)

    # Finally write the output file
    wavwrite(transpose(output), args.outfile, sampleRate)

    def _analyse(self, filepath):
        audio = to_mono(wavread(filepath)[0])
        audio = audio.astype('float32')
        
        w = Windowing(type = 'hann')
        fft = FFT() # this gives us a complex FFT
        c2p = CartesianToPolar() # and this turns it into a pair (magnitude, phase)
        hfc_detect = OnsetDetection(method = 'hfc')
        complex_detect = OnsetDetection(method = 'complex')
        rms_detect = RMS()
        spec = Spectrum()
        #pd = PitchDetection()
        flux = Flux()
        pool = Pool()
        #wap = WarpedAutoCorrelation()
        
    
        # let's get down to business
        print 'Computing onset detection functions...'
        for frame in FrameGenerator(audio, frameSize = self.frame_size,\
            hopSize = self.hop_size):
            mag, phase, = c2p(fft(w(frame)))
            spectrum = spec(w(frame))
            f = flux(spectrum)
            #pitch = pd(spectrum)
            pool.add('hfc', hfc_detect(mag, phase))
            pool.add('complex', complex_detect(mag, phase))
            pool.add('rms', rms_detect(frame))
            pool.add('flux', f)
            #pool.add('pitch', pitch[0])
        #print pool['pitch']
        #pool.add('autoc', wap(pool['pitch']))
     

        return pool, audio

    def estimate_f0s(self, audio_path):
        if not os.path.exists(audio_path):
            raise ValueError('Invalid audio path')

        x, fs, _ = wavread(audio_path)

        # make x mono if stereo
        if x.ndim > 1:
            _, n_channels = x.shape
            x = x.sum(axis=1)/n_channels

        X = self._stft(x, fs)

        # Section 2.1 Spectrally whiten the signal to suppress timbral information
        Y = self._spectral_whitening(X, fs)

        # perform iterative estimation of the fundamental periods in the audio file
        f0_estimations = self._iterative_est(Y, fs)
        
        # get notes which correspond to these frequency estimates
        notes = []
        for frame_ests in f0_estimations:
            notes.append([self._freq_to_note(f) for f in frame_ests])

        return f0_estimations, notes

def loadFiles(path):
	"""reads wave files from path and returns dictionary with fields:
        - "name" - name of file
        - "nameGender" - a sex readed from filename
        - "signal" - numpy array with sound signal readed from file
        - "sampleRate" - sample rate of the file

        and dictionary that contains numbers of male and female voices
	"""
	print "reading files..."

	files = [ f for f in listdir(path) if isfile(join(path,f)) and splitext(f)[1] == ".wav" ]

	samples = []
	maleCount = 0
	femaleCount = 0
	for f in files:
		p = path + '/' + f

		print "...", f
		data,rate,encoding=wavread(p)
		sig=[mean(d) for d in data]    
		samples.append({'name': f, 'nameGender': f[-5:-4], 'signal': sig, 'sampleRate': rate})
        
		if f[-5:-4] == "M":
			maleCount += 1
		else:
			femaleCount += 1
    
	counters = {"maleCount":maleCount, "femaleCount":femaleCount}
	return samples, counters

 def open_wav_audiolab(self, filename):
     #http://scikits.appspot.com/audiolab
     from scikits.audiolab import wavread
     results, sample_frequency,encoding = wavread(filename)
     self.sample_rate = sample_frequency
     print 'Sample Rate is ', sample_frequency
     return results, self.sample_rate

def normalize_target_audio(input_file='moviehires_endpos_beta02.imatsh.wav', 
                           sources_expr='/home/mkc/Music/GoldbergVariations/*48_1.wav', write_me=False, amp_factor=0.5, proc_audio=True):
    """
    Per-variation normalization of concatenated imatsh file using individual sources as locators
    Assumes that the input_file and the source_dir have the same sample rate
    inputs:
        input_file  - the file to be processed (locally normalized)
        sources_expr- regular expression for input files
        write_me    - write output files when true [False]
        amp_factor  - amplitude change factor (proportion of full scale normalization) [0.5]
        proc_audio  - whether to process target audio using source audio info [1]
    outputs:
        sample_locators - sample locators for each variation
        audio_summaries - min, max, rms values for each variation        
    output files:
        output_file = {input_file_stem}+'norm.'+{input_ext}
    """
    # Compute min, max, rms per source file
    flist = glob.glob(sources_expr)
    flist.sort()
    sample_locators = [0]
    audio_summaries = []
    ext_pos = input_file.rindex('.')
    outfile_stem, ext = input_file[:ext_pos], input_file[ext_pos+1:]
    for i,f in enumerate(flist):
        x,sr,fmt = skaud.wavread(f)
        print f, sr, fmt
        if(len(x.shape)>1):
            x = x[:,0] # Take left-channel only
        sample_locators.extend([len(x)])
        audio_summaries.append([max(abs(x)), np.sqrt(np.mean(x**2))])
        if proc_audio:
            y,sr_y,fmt_y = skaud.wavread(input_file, first=np.cumsum(sample_locators)[-2], last=sample_locators[-1])
            if sr != sr_y:
                raise ValueError("input and source sample rates don't match: %d,%d"%(sr,sr_y))
            audio_summaries.append([max(abs(y[:,0])), np.sqrt(np.mean(y[:,0]**2))])
            max_val = audio_summaries[-1][0]
            rms_val = audio_summaries[-1][1]
            norm_cf = amp_factor / max_val + (1 - amp_factor)
            outfile = outfile_stem+'_%02d.%s'%(i+1,ext)
            max_amp_val = norm_cf * max_val
            rms_amp_val = norm_cf * rms_val
            print '%s: nrm=%05.2fdB, peak=%05.2fdB, *peak=%05.2fdB, rms=%05.2fdB, *rms=%05.2fdB'%(
                outfile, dB(norm_cf), dB(max_val), dB(max_amp_val), dB(rms_val), dB(rms_amp_val))
            if(write_me):
                skaud.wavwrite(norm_cf*y, outfile, sr, fmt)
    return np.cumsum(sample_locators), np.array(audio_summaries)