o Met,@spddlZddlmZddlmZddlZddlmZ d-deeefdedeeeffd d Z d-d eeefdedeeeffd d Z     d.de dededede def ddZ d/de de de defddZ      d0de de de dedeededee efde defddZ ! "d1d#ed$ed%ed&eedef d'd(Z ) d2d*e de dee de def d+d,ZdS)3N)Optional)Union)TensorFfreqhtkreturnc Cs|rt|trdtd|dSdtd|dSd}d}|||}d}|||}tdd}t|tr\|t||d |}||k|j} || |d | }|S||krm|t||d |}|S) aConvert Hz to Mels. Args: freq (Union[Tensor, float]): The input tensor with arbitrary shape. htk (bool, optional): Use htk scaling. Defaults to False. Returns: Union[Tensor, float]: Frequency in mels. Examples: .. code-block:: python import paddle val = 3.0 htk_flag = True mel_paddle_tensor = paddle.audio.functional.hz_to_mel( paddle.to_tensor(val), htk_flag) F@?@竪P@@@皙@;@绽|=) isinstancerpaddlelog10mathlogastypedtype) rrf_minf_spmels min_log_hz min_log_mellogsteptargetmaskr!RD:\Projects\ConvertPro\env\Lib\site-packages\paddle/audio/functional/functional.py hz_to_mels0     r#melc Cs|r dd|ddSd}d}|||}d}|||}tdd }t|trI|t|||}||k|j} || |d | }|S||krX|t|||}|S) a Convert mel bin numbers to frequencies. Args: mel (Union[float, Tensor]): The mel frequency represented as a tensor with arbitrary shape. htk (bool, optional): Use htk scaling. Defaults to False. Returns: Union[float, Tensor]: Frequencies in Hz. Examples: .. code-block:: python import paddle val = 3.0 htk_flag = True mel_paddle_tensor = paddle.audio.functional.mel_to_hz( paddle.to_tensor(val), htk_flag) r $@rr r r r rrr)rrrrrexprr) r$rrrfreqsrrrrr r!r!r" mel_to_hzLs$   r(@r @float32n_melsrf_maxrc Cs:t||d}t||d}tj||||d}t||d}|S)a2Compute mel frequencies. Args: n_mels (int, optional): Number of mel bins. Defaults to 64. f_min (float, optional): Minimum frequency in Hz. Defaults to 0.0. fmax (float, optional): Maximum frequency in Hz. Defaults to 11025.0. htk (bool, optional): Use htk scaling. Defaults to False. dtype (str, optional): The data type of the return frequencies. Defaults to 'float32'. Returns: Tensor: Tensor of n_mels frequencies in Hz with shape `(n_mels,)`. Examples: .. code-block:: python import paddle n_mels = 64 f_min = 0.5 f_max = 10000 htk_flag = True paddle_mel_freq = paddle.audio.functional.mel_frequencies( n_mels, f_min, f_max, htk_flag, 'float64') )rr)r#rlinspacer() r,rr-rrZmin_melZmax_melrr'r!r!r"mel_frequenciesws   r0srn_fftcCs&tjdt|dtd|d|dS)aCompute fourier frequencies. Args: sr (int): Sample rate. n_fft (int): Number of fft bins. dtype (str, optional): The data type of the return frequencies. Defaults to 'float32'. Returns: Tensor: FFT frequencies in Hz with shape `(n_fft//2 + 1,)`. Examples: .. code-block:: python import paddle sr = 16000 n_fft = 128 fft_freq = paddle.audio.functional.fft_frequencies(sr, n_fft) rrr.)rr/floatintr1r2rr!r!r"fft_frequenciess&r7slaneynormcCs>|dur t|d}tj|td|df|d}t|||d} t|d||||d} | dd| dd} | d| d} t|D]&} | |  | | }| | d| | d}tt |t |||| <qE|d krd | d|d| d|}||d9}|St |tst |trtj j j||dd }|S) aeCompute fbank matrix. Args: sr (int): Sample rate. n_fft (int): Number of fft bins. n_mels (int, optional): Number of mel bins. Defaults to 64. f_min (float, optional): Minimum frequency in Hz. Defaults to 0.0. f_max (Optional[float], optional): Maximum frequency in Hz. Defaults to None. htk (bool, optional): Use htk scaling. Defaults to False. norm (Union[str, float], optional): Type of normalization. Defaults to 'slaney'. dtype (str, optional): The data type of the return matrix. Defaults to 'float32'. Returns: Tensor: Mel transform matrix with shape `(n_mels, n_fft//2 + 1)`. Examples: .. code-block:: python import paddle n_mfcc = 23 n_mels = 51 paddle_dct = paddle.audio.functional.create_dct(n_mfcc, n_mels) Nr3rr.r6)rr-rrrr8@)pZaxis)r4rZzerosr5r7r0 unsqueezerangemaximumZ zeros_likeminimumrnnZ functional normalize)r1r2r,rr-rr9rweightsZfftfreqsZmel_fZfdiffZrampsilowerupperZenormr!r!r"compute_fbank_matrixs2!      rGr rT@spect ref_valueamintop_dbcCs|dkrtd|dkrtdt|}dtt|||}|dtt||8}|durF|dkr:tdt||||}|S)aXConvert a power spectrogram (amplitude squared) to decibel (dB) units. The function computes the scaling `10 * log10(x / ref)` in a numerically stable way. Args: spect (Tensor): STFT power spectrogram. 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): Minimum threshold. Defaults to 1e-10. top_db (Optional[float], optional): Threshold the output at `top_db` below the peak. Defaults to None. Returns: Tensor: Power spectrogram in db scale. Examples: .. code-block:: python import paddle val = 3.0 decibel_paddle = paddle.audio.functional.power_to_db( paddle.to_tensor(val)) rzamin must be strictly positivez#ref_value must be strictly positiver%Nztop_db must be non-negative) ExceptionrZ ones_likerr?rmax)rIrJrKrLZonesZlog_specr!r!r" power_to_dbs rOorthon_mfcccCstj||d}tj||dd}ttjt||d|}|dur,|d9}|jS|dks2J|ddtd9<|tdt|9}|jS) alCreate a discrete cosine transform(DCT) matrix. Args: n_mfcc (int): Number of mel frequency cepstral coefficients. n_mels (int): Number of mel filterbanks. norm (Optional[str], optional): Normalizaiton type. Defaults to 'ortho'. dtype (str, optional): The data type of the return matrix. Defaults to 'float32'. Returns: Tensor: The DCT matrix with shape `(n_mels, n_mfcc)`. Examples: .. code-block:: python import paddle n_mfcc = 23 n_mels = 257 dct = paddle.audio.functional.create_dct(n_mfcc, n_mels) r.rg?Nr;rPrr ) rZaranger=cosrpir4sqrtT)rQr,r9rnkdctr!r!r" create_dct&s rY)F)r)r r*Fr+)r+)r)r NFr8r+)r rrH)rPr+)rtypingrrrrr4boolr#r(r5strr0r7rGrOrYr!r!r!r"s     6  + &  I ,