o MeO@sddlmZddlmZddlmZddlZddlmZddlmZddl m Z ddl m Z dd l m Z dd l mZGd d d ejZGd ddejZGdddejZGdddejZdS))partial)Optional)UnionN)Tensor)compute_fbank_matrix) create_dct) power_to_db) get_windowcsreZdZdZ        dd ed eed eed ed ededededdffdd Z de de fddZ Z S) SpectrogramaaCompute spectrogram of given signals, typically audio waveforms. The spectorgram is defined as the complex norm of the short-time Fourier transformation. Args: n_fft (int, optional): The number of frequency components of the discrete Fourier transform. Defaults to 512. hop_length (Optional[int], optional): The hop length of the short time FFT. If `None`, it is set to `win_length//4`. Defaults to None. win_length (Optional[int], optional): The window length of the short time FFT. If `None`, it is set to same as `n_fft`. Defaults to None. window (str, optional): The window function applied to the signal before the Fourier transform. Supported window functions: 'hamming', 'hann', 'kaiser', 'gaussian', 'exponential', 'triang', 'bohman', 'blackman', 'cosine', 'tukey', 'taylor'. Defaults to 'hann'. power (float, optional): Exponent for the magnitude spectrogram. Defaults to 2.0. center (bool, optional): Whether to pad `x` to make that the :math:`t imes hop\_length` at the center of `t`-th frame. Defaults to True. pad_mode (str, optional): Choose padding pattern when `center` is `True`. Defaults to 'reflect'. dtype (str, optional): Data type of input and window. Defaults to 'float32'. Returns: :ref:`api_paddle_nn_Layer`. An instance of Spectrogram. Examples: .. code-block:: python import paddle from paddle.audio.features import Spectrogram sample_rate = 16000 wav_duration = 0.5 num_channels = 1 num_frames = sample_rate * wav_duration wav_data = paddle.linspace(-1.0, 1.0, num_frames) * 0.1 waveform = wav_data.tile([num_channels, 1]) feature_extractor = Spectrogram(n_fft=512, window = 'hann', power = 1.0) feats = feature_extractor(waveform) Nhann?Treflectfloat32n_fft hop_length win_lengthwindowpowercenterpad_modedtypereturnc srtt||dksJd||_|dur|}t||d|d|_ttjj ||||j||d|_ | d|jdS)Nrz!Power of spectrogram must be > 0.T)Zfftbinsr)rrrrrr fft_window) superr __init__rr rrpaddlesignalstft_stftregister_buffer) selfrrrrrrrr __class__LD:\Projects\ConvertPro\env\Lib\site-packages\paddle/audio/features/layers.pyr@s& zSpectrogram.__init__xcCs"||}tt||j}|S)z Args: x (Tensor): Tensor of waveforms with shape `(N, T)` Returns: Tensor: Spectrograms with shape `(N, n_fft//2 + 1, num_frames)`. )r rpowabsr)r"r'rZ spectrogramr%r%r&forward^s zSpectrogram.forward)r r Nr rTrr) __name__ __module__ __qualname____doc__intrstrfloatboolrrr* __classcell__r%r%r#r&r s<$r cseZdZdZ         d#dededeedeedededededededeedede eefdeddffdd Z d e de fd!d"Z Z S)$MelSpectrograma Compute the melspectrogram of given signals, typically audio waveforms. It is computed by multiplying spectrogram with Mel filter bank matrix. Args: sr (int, optional): Sample rate. Defaults to 22050. n_fft (int, optional): The number of frequency components of the discrete Fourier transform. Defaults to 512. hop_length (Optional[int], optional): The hop length of the short time FFT. If `None`, it is set to `win_length//4`. Defaults to None. win_length (Optional[int], optional): The window length of the short time FFT. If `None`, it is set to same as `n_fft`. Defaults to None. window (str, optional): The window function applied to the signal before the Fourier transform. Supported window functions: 'hamming', 'hann', 'kaiser', 'gaussian', 'exponential', 'triang', 'bohman', 'blackman', 'cosine', 'tukey', 'taylor'. Defaults to 'hann'. power (float, optional): Exponent for the magnitude spectrogram. Defaults to 2.0. center (bool, optional): Whether to pad `x` to make that the :math:`t imes hop\_length` at the center of `t`-th frame. Defaults to True. pad_mode (str, optional): Choose padding pattern when `center` is `True`. Defaults to 'reflect'. n_mels (int, optional): Number of mel bins. Defaults to 64. f_min (float, optional): Minimum frequency in Hz. Defaults to 50.0. f_max (Optional[float], optional): Maximum frequency in Hz. Defaults to None. htk (bool, optional): Use HTK formula in computing fbank matrix. Defaults to False. norm (Union[str, float], optional): Type of normalization in computing fbank matrix. Slaney-style is used by default. You can specify norm=1.0/2.0 to use customized p-norm normalization. Defaults to 'slaney'. dtype (str, optional): Data type of input and window. Defaults to 'float32'. Returns: :ref:`api_paddle_nn_Layer`. An instance of MelSpectrogram. Examples: .. code-block:: python import paddle from paddle.audio.features import MelSpectrogram sample_rate = 16000 wav_duration = 0.5 num_channels = 1 num_frames = sample_rate * wav_duration wav_data = paddle.linspace(-1.0, 1.0, num_frames) * 0.1 waveform = wav_data.tile([num_channels, 1]) feature_extractor = MelSpectrogram(sr=sample_rate, n_fft=512, window = 'hann', power = 1.0) feats = feature_extractor(waveform) "Vr Nr @Tr@I@Fslaneyrsrrrrrrrrn_melsf_minf_maxhtknormrrc stt|t||||||||d|_| |_| |_| |_| |_| |_ | dur+|d} t ||| | | | | |d|_ | d|j dS)N)rrrrrrrrr)r;rr<r=r>r?r@r fbank_matrix) rr4rr _spectrogramr<r=r>r?r@rrAr!)r"r;rrrrrrrr<r=r>r?r@rr#r%r&rs6zMelSpectrogram.__init__r'cCs||}t|j|}|S)z Args: x (Tensor): Tensor of waveforms with shape `(N, T)` Returns: Tensor: Mel spectrograms with shape `(N, n_mels, num_frames)`. )rBrmatmulrA)r"r'Z spect_feature mel_featurer%r%r&r*s zMelSpectrogram.forward)r5r6r Nr r7Trr8r9NFr:rr+r,r-r.r/rr0r1r2rrrr*r3r%r%r#r&r4ks`'     *r4c%seZdZdZ            d'dededeedeedededededededeedede eefdededeed ed!df$fd"d# Z d$e d!e fd%d&Z Z S)(LogMelSpectrograma Compute log-mel-spectrogram feature of given signals, typically audio waveforms. Args: sr (int, optional): Sample rate. Defaults to 22050. n_fft (int, optional): The number of frequency components of the discrete Fourier transform. Defaults to 512. hop_length (Optional[int], optional): The hop length of the short time FFT. If `None`, it is set to `win_length//4`. Defaults to None. win_length (Optional[int], optional): The window length of the short time FFT. If `None`, it is set to same as `n_fft`. Defaults to None. window (str, optional): The window function applied to the signal before the Fourier transform. Supported window functions: 'hamming', 'hann', 'kaiser', 'gaussian', 'exponential', 'triang', 'bohman', 'blackman', 'cosine', 'tukey', 'taylor'. Defaults to 'hann'. power (float, optional): Exponent for the magnitude spectrogram. Defaults to 2.0. center (bool, optional): Whether to pad `x` to make that the :math:`t imes hop\_length` at the center of `t`-th frame. Defaults to True. pad_mode (str, optional): Choose padding pattern when `center` is `True`. Defaults to 'reflect'. n_mels (int, optional): Number of mel bins. Defaults to 64. f_min (float, optional): Minimum frequency in Hz. Defaults to 50.0. f_max (Optional[float], optional): Maximum frequency in Hz. Defaults to None. htk (bool, optional): Use HTK formula in computing fbank matrix. Defaults to False. norm (Union[str, float], optional): Type of normalization in computing fbank matrix. Slaney-style is used by default. You can specify norm=1.0/2.0 to use customized p-norm normalization. Defaults to 'slaney'. ref_value (float, optional): The reference value. If smaller than 1.0, the db level of the signal will be pulled up accordingly. Otherwise, the db level is pushed down. Defaults to 1.0. amin (float, optional): The minimum value of input magnitude. Defaults to 1e-10. top_db (Optional[float], optional): The maximum db value of spectrogram. Defaults to None. dtype (str, optional): Data type of input and window. Defaults to 'float32'. Returns: :ref:`api_paddle_nn_Layer`. An instance of LogMelSpectrogram. Examples: .. code-block:: python import paddle from paddle.audio.features import LogMelSpectrogram sample_rate = 16000 wav_duration = 0.5 num_channels = 1 num_frames = sample_rate * wav_duration wav_data = paddle.linspace(-1.0, 1.0, num_frames) * 0.1 waveform = wav_data.tile([num_channels, 1]) feature_extractor = LogMelSpectrogram(sr=sample_rate, n_fft=512, window = 'hann', power = 1.0) feats = feature_extractor(waveform) r5r Nr r7Trr8r9Fr:r绽|=rr;rrrrrrrr<r=r>r?r@ ref_valueamintop_dbrrcsJtt|t||||||||| | | | | |d|_||_||_||_dS)N)r;rrrrrrrr<r=r>r?r@r)rrFrr4_melspectrogramrHrIrJ)r"r;rrrrrrrr<r=r>r?r@rHrIrJrr#r%r&rs& zLogMelSpectrogram.__init__r'cCs$||}t||j|j|jd}|S)z Args: x (Tensor): Tensor of waveforms with shape `(N, T)` Returns: Tensor: Log mel spectrograms with shape `(N, n_mels, num_frames)`. )rHrIrJ)rKr rHrIrJ)r"r'rDlog_mel_featurer%r%r&r*s zLogMelSpectrogram.forward)r5r NNr r7Trr8r9NFr:rrGNrrEr%r%r#r&rFsr*     'rFc'seZdZdZ             d)dedededeedeedededededededeedede eefded ed!eed"ed#df&fd$d% Z d&e d#e fd'd(Z Z S)*MFCCa Compute mel frequency cepstral coefficients(MFCCs) feature of given waveforms. Args: sr (int, optional): Sample rate. Defaults to 22050. n_mfcc (int, optional): [description]. Defaults to 40. n_fft (int, optional): The number of frequency components of the discrete Fourier transform. Defaults to 512. hop_length (Optional[int], optional): The hop length of the short time FFT. If `None`, it is set to `win_length//4`. Defaults to None. win_length (Optional[int], optional): The window length of the short time FFT. If `None`, it is set to same as `n_fft`. Defaults to None. window (str, optional): The window function applied to the signal before the Fourier transform. Supported window functions: 'hamming', 'hann', 'kaiser', 'gaussian', 'exponential', 'triang', 'bohman', 'blackman', 'cosine', 'tukey', 'taylor'. Defaults to 'hann'. power (float, optional): Exponent for the magnitude spectrogram. Defaults to 2.0. center (bool, optional): Whether to pad `x` to make that the :math:`t imes hop\_length` at the center of `t`-th frame. Defaults to True. pad_mode (str, optional): Choose padding pattern when `center` is `True`. Defaults to 'reflect'. n_mels (int, optional): Number of mel bins. Defaults to 64. f_min (float, optional): Minimum frequency in Hz. Defaults to 50.0. f_max (Optional[float], optional): Maximum frequency in Hz. Defaults to None. htk (bool, optional): Use HTK formula in computing fbank matrix. Defaults to False. norm (Union[str, float], optional): Type of normalization in computing fbank matrix. Slaney-style is used by default. You can specify norm=1.0/2.0 to use customized p-norm normalization. Defaults to 'slaney'. ref_value (float, optional): The reference value. If smaller than 1.0, the db level of the signal will be pulled up accordingly. Otherwise, the db level is pushed down. Defaults to 1.0. amin (float, optional): The minimum value of input magnitude. Defaults to 1e-10. top_db (Optional[float], optional): The maximum db value of spectrogram. Defaults to None. dtype (str, optional): Data type of input and window. Defaults to 'float32'. Returns: :ref:`api_paddle_nn_Layer`. An instance of MFCC. Examples: .. code-block:: python import paddle from paddle.audio.features import MFCC sample_rate = 16000 wav_duration = 0.5 num_channels = 1 num_frames = sample_rate * wav_duration wav_data = paddle.linspace(-1.0, 1.0, num_frames) * 0.1 waveform = wav_data.tile([num_channels, 1]) feature_extractor = MFCC(sr=sample_rate, n_fft=512, window = 'hann') feats = feature_extractor(waveform) r5(r Nr r7Trr8r9Fr:rrGrr;n_mfccrrrrrrrr<r=r>r?r@rHrIrJrrcstt||| ksJd|| ftdid|d|d|d|d|d|d|d | d | d | d | d | d|d|d|d|d||_t|| |d|_|d|jdS)Nz-n_mfcc cannot be larger than n_mels: %d vs %dr;rrrrrrrr<r=r>r?r@rHrIrJr)rOr<r dct_matrixr%)rrMrrF_log_melspectrogramrrPr!)r"r;rOrrrrrrrr<r=r>r?r@rHrIrJrr#r%r&rUsN      z MFCC.__init__r'cCs(||}t|d|jd}|S)z Args: x (Tensor): Tensor of waveforms with shape `(N, T)` Returns: Tensor: Mel frequency cepstral coefficients with shape `(N, n_mfcc, num_frames)`. )rr)rQrrCZ transposerP)r"r'rLZmfccr%r%r&r*s z MFCC.forward)r5rNr NNr r7Trr8r9NFr:rrGNrrEr%r%r#r&rM*sx+      *rM) functoolsrtypingrrrZ paddle.nnnnrZ functionalrrr Zfunctional.windowr ZLayerr r4rFrMr%r%r%r&s        O^a