源码类别
热门源码
 当前位置：源码 > Python源码 >

Python源码示例：hparams.hparams.fft_size()

示例1
def _stft(y):
    return librosa.stft(y=y, n_fft=hparams.fft_size, hop_length=get_hop_size())
示例2
def _stft(y):
	return librosa.stft(y=y, n_fft=hparams.fft_size, hop_length=get_hop_size())
示例3
def _build_mel_basis():
	assert hparams.fmax <= hparams.sample_rate // 2
	return librosa.filters.mel(hparams.sample_rate, hparams.fft_size, n_mels=hparams.num_mels,
							   fmin=hparams.fmin, fmax=hparams.fmax)
示例4
def _stft(y):
	return librosa.stft(y=y, n_fft=hparams.fft_size, hop_length=get_hop_size())
示例5
def _build_mel_basis():
	assert hparams.fmax <= hparams.sample_rate // 2
	return librosa.filters.mel(hparams.sample_rate, hparams.fft_size, n_mels=hparams.num_mels,
							   fmin=hparams.fmin, fmax=hparams.fmax)
示例6
def _lws_processor():
    return lws.lws(hparams.fft_size, hparams.hop_size, mode="speech")


# Conversions:
示例7
def _build_mel_basis():
    if hparams.fmax is not None:
        assert hparams.fmax <= hparams.sample_rate // 2
    return librosa.filters.mel(hparams.sample_rate, hparams.fft_size,
                               fmin=hparams.fmin, fmax=hparams.fmax,
                               n_mels=hparams.num_mels)
示例8
def _stft(y):
	return librosa.stft(y=y, n_fft=hparams.fft_size, hop_length=get_hop_size())
示例9
def _build_mel_basis():
	assert hparams.fmax <= hparams.sample_rate // 2
	return librosa.filters.mel(hparams.sample_rate, hparams.fft_size, n_mels=hparams.num_mels,
							   fmin=hparams.fmin, fmax=hparams.fmax)
示例10
def _stft(y):
	return librosa.stft(y=y, n_fft=hparams.fft_size, hop_length=get_hop_size())
示例11
def _build_mel_basis():
	assert hparams.fmax <= hparams.sample_rate // 2
	return librosa.filters.mel(hparams.sample_rate, hparams.fft_size, n_mels=hparams.num_mels,
							   fmin=hparams.fmin, fmax=hparams.fmax)
示例12
def _lws_processor():
    return lws.lws(hparams.fft_size, get_hop_size(), mode="speech")
示例13
def _build_mel_basis():
    assert hparams.fmax <= hparams.sample_rate // 2
    return librosa.filters.mel(hparams.sample_rate, hparams.fft_size,
                               fmin=hparams.fmin, fmax=hparams.fmax,
                               n_mels=hparams.num_mels)
示例14
def _lws_processor():
    return lws.lws(hparams.fft_size, get_hop_size(), mode="speech")
示例15
def _build_mel_basis():
    assert hparams.fmax <= hparams.sample_rate // 2
    return librosa.filters.mel(hparams.sample_rate, hparams.fft_size,
                               fmin=hparams.fmin, fmax=hparams.fmax,
                               n_mels=hparams.num_mels)
示例16
def _build_mel_basis():
    assert hparams.fmax <= hparams.sample_rate // 2
    return librosa.filters.mel(hparams.sample_rate, hparams.fft_size,
                               fmin=hparams.fmin, fmax=hparams.fmax,
                               n_mels=hparams.num_mels)
示例17
def _lws_processor():
    return lws.lws(hparams.fft_size, get_hop_size(), mode="speech")
示例18
def _build_mel_basis():
    assert hparams.fmax <= hparams.sample_rate // 2
    return librosa.filters.mel(hparams.sample_rate, hparams.fft_size,
                               fmin=hparams.fmin, fmax=hparams.fmax,
                               n_mels=hparams.num_mels)
示例19
def _process_utterance(out_dir, index, wav_path, text):
    # Load the audio to a numpy array:
    wav = audio.load_wav(wav_path)

    if hparams.rescaling:
        wav = wav / np.abs(wav).max() * hparams.rescaling_max

    # Mu-law quantize
    if is_mulaw_quantize(hparams.input_type):
        # [0, quantize_channels)
        out = P.mulaw_quantize(wav, hparams.quantize_channels)

        # Trim silences
        start, end = audio.start_and_end_indices(out, hparams.silence_threshold)
        wav = wav[start:end]
        out = out[start:end]
        constant_values = P.mulaw_quantize(0, hparams.quantize_channels)
        out_dtype = np.int16
    elif is_mulaw(hparams.input_type):
        # [-1, 1]
        out = P.mulaw(wav, hparams.quantize_channels)
        constant_values = P.mulaw(0.0, hparams.quantize_channels)
        out_dtype = np.float32
    else:
        # [-1, 1]
        out = wav
        constant_values = 0.0
        out_dtype = np.float32

    # Compute a mel-scale spectrogram from the trimmed wav:
    # (N, D)
    mel_spectrogram = audio.melspectrogram(wav).astype(np.float32).T
    # lws pads zeros internally before performing stft
    # this is needed to adjust time resolution between audio and mel-spectrogram
    l, r = audio.lws_pad_lr(wav, hparams.fft_size, audio.get_hop_size())

    # zero pad for quantized signal
    out = np.pad(out, (l, r), mode="constant", constant_values=constant_values)
    N = mel_spectrogram.shape[0]
    assert len(out) >= N * audio.get_hop_size()

    # time resolution adjustment
    # ensure length of raw audio is multiple of hop_size so that we can use
    # transposed convolution to upsample
    out = out[:N * audio.get_hop_size()]
    assert len(out) % audio.get_hop_size() == 0

    timesteps = len(out)

    # Write the spectrograms to disk:
    audio_filename = 'ljspeech-audio-%05d.npy' % index
    mel_filename = 'ljspeech-mel-%05d.npy' % index
    np.save(os.path.join(out_dir, audio_filename),
            out.astype(out_dtype), allow_pickle=False)
    np.save(os.path.join(out_dir, mel_filename),
            mel_spectrogram.astype(np.float32), allow_pickle=False)

    # Return a tuple describing this training example:
    return (audio_filename, mel_filename, timesteps, text)
示例20
def _process_utterance(out_dir, index, wav_path, text):
    # Load the audio to a numpy array:
    wav = audio.load_wav(wav_path)

    if hparams.rescaling:
        wav = wav / np.abs(wav).max() * hparams.rescaling_max

    # Mu-law quantize
    if is_mulaw_quantize(hparams.input_type):
        # [0, quantize_channels)
        out = P.mulaw_quantize(wav, hparams.quantize_channels)

        # Trim silences
        start, end = audio.start_and_end_indices(out, hparams.silence_threshold)
        wav = wav[start:end]
        out = out[start:end]
        constant_values = P.mulaw_quantize(0, hparams.quantize_channels)
        out_dtype = np.int16
    elif is_mulaw(hparams.input_type):
        # [-1, 1]
        out = P.mulaw(wav, hparams.quantize_channels)
        constant_values = P.mulaw(0.0, hparams.quantize_channels)
        out_dtype = np.float32
    else:
        # [-1, 1]
        out = wav
        constant_values = 0.0
        out_dtype = np.float32

    # Compute a mel-scale spectrogram from the trimmed wav:
    # (N, D)
    mel_spectrogram = audio.melspectrogram(wav).astype(np.float32).T
    # lws pads zeros internally before performing stft
    # this is needed to adjust time resolution between audio and mel-spectrogram
    l, r = audio.lws_pad_lr(wav, hparams.fft_size, audio.get_hop_size())

    # zero pad for quantized signal
    out = np.pad(out, (l, r), mode="constant", constant_values=constant_values)
    N = mel_spectrogram.shape[0]
    assert len(out) >= N * audio.get_hop_size()

    # time resolution adjustment
    # ensure length of raw audio is multiple of hop_size so that we can use
    # transposed convolution to upsample
    out = out[:N * audio.get_hop_size()]
    assert len(out) % audio.get_hop_size() == 0

    timesteps = len(out)

    # Write the spectrograms to disk:
    audio_filename = 'ljspeech-audio-%05d.npy' % index
    mel_filename = 'ljspeech-mel-%05d.npy' % index
    np.save(os.path.join(out_dir, audio_filename),
            out.astype(out_dtype), allow_pickle=False)
    np.save(os.path.join(out_dir, mel_filename),
            mel_spectrogram.astype(np.float32), allow_pickle=False)

    # Return a tuple describing this training example:
    return (audio_filename, mel_filename, timesteps, text)