o 8wi)W@sdZddlZddlZddlmZd&ddZd'd d Zd(d dZd)ddZ    d*ddZ d+ddZ ddZ d,ddZ ddZ d-d d!Zd"d#Zd$d%ZdS).z Low level signal processing utilities Authors * Peter Plantinga 2020 * Francois Grondin 2020 * William Aris 2020 * Samuele Cornell 2020 * Sarthak Yadav 2022 N)versionavglinearcCslt|jdkr |d}|dvsJ|dvsJ|dkrN|dur,tjt|ddd}nrtjt|ddd }t|jd krIt|tjrI|d }||}nP|d kr|durdt tj|d ddd}n:tjt |d ddd }t|jd krt|tjr|d }t ||}n|d krtj t|dddd}nt |dkr|S|dkrtj dt|ddSt )aQCompute amplitude of a batch of waveforms. Arguments --------- waveforms : tensor The waveforms used for computing amplitude. Shape should be `[time]` or `[batch, time]` or `[batch, time, channels]`. lengths : tensor The lengths of the waveforms excluding the padding. Shape should be a single dimension, `[batch]`. amp_type : str Whether to compute "avg" average or "peak" amplitude. Choose between ["avg", "peak"]. scale : str Whether to compute amplitude in "dB" or "linear" scale. Choose between ["linear", "dB"]. Returns ------- The average amplitude of the waveforms. Example ------- >>> signal = torch.sin(torch.arange(16000.0)).unsqueeze(0) >>> compute_amplitude(signal, signal.size(1)) tensor([[0.6366]]) r)rrmspeakrdBrNT)dimkeepdim)inputr r rrrr i)min)lenshape unsqueezetorchmeanabssum isinstanceTensorsqrtpowmaxNotImplementedErrorclamplog10) waveformslengthsamp_typescaleoutwav_sumr&e/home/ubuntu/sommelier/.venv/lib/python3.10/site-packages/speechbrain/processing/signal_processing.pycompute_amplitudes8      r(+=cCsR|dvsJd}t|jdkrd}|d}t||||}|r%|d}||S)aThis function normalizes a signal to unitary average or peak amplitude. Arguments --------- waveforms : tensor The waveforms to normalize. Shape should be `[batch, time]` or `[batch, time, channels]`. lengths : tensor The lengths of the waveforms excluding the padding. Shape should be a single dimension, `[batch]`. amp_type : str Whether one wants to normalize with respect to "avg" or "peak" amplitude. Choose between ["avg", "peak"]. Note: for "avg" clipping is not prevented and can occur. eps : float A small number to add to the denominator to prevent NaN. Returns ------- waveforms : tensor Normalized level waveform. )rrFrTr)rrrr(squeeze)r r!r"eps batch_addeddenr&r&r' normalizeVs   r.rư>cCs0|j|dd}|j|dd}||||}|S)aThis function normalizes the mean and std of the input waveform (along the specified axis). Arguments --------- waveforms : tensor The waveforms to normalize. Shape should be `[batch, time]` or `[batch, time, channels]`. dims : int or tuple The dimension(s) on which mean and std are computed eps : float A small number to add to the denominator to prevent NaN. Returns ------- waveforms : tensor Normalized level waveform. T)r )rstd)r dimsr+rr0r&r&r' mean_std_normzsr2cCs|dvsJ|dvs Jd}t|jdkrd}|d}t|||}|dkr+||}n|dkr6t||}ntd |rA|d}|S) aThis functions performs signal rescaling to a target level. Arguments --------- waveforms : tensor The waveforms to normalize. Shape should be `[batch, time]` or `[batch, time, channels]`. lengths : tensor The lengths of the waveforms excluding the padding. Shape should be a single dimension, `[batch]`. target_lvl : float Target lvl in dB or linear scale. amp_type : str Whether one wants to rescale with respect to "avg" or "peak" amplitude. Choose between ["avg", "peak"]. scale : str whether target_lvl belongs to linear or dB scale. Choose between ["linear", "dB"]. Returns ------- waveforms : tensor Rescaled waveforms. )rrrFrTrrr z+Invalid scale, choose between dB and linear)rrrr.dB_to_amplituderr*)r r! target_lvlr"r#r,r$r&r&r'rescales      r5constantFcCst|jdkr td|dd}|dd}t|tr&tjjj |||d}|r| d| d}|dkr@|dd |f}d}tj | d| d||j d } |d|d f} |dd |f} tj | | | fdd }ttjtd krdd lm} | || |} | j| | dd }nOt|d}t|d}|d\}}|d\}}tj||||||||gdd }tj|d| dgd}ntjjj||||t|ts|ndd}|ddS)aMUse torch.nn.functional to perform 1d padding and conv. Arguments --------- waveform : tensor The tensor to perform operations on. kernel : tensor The filter to apply during convolution. padding : int or tuple The padding (pad_left, pad_right) to apply. If an integer is passed instead, this is passed to the conv1d function and pad_type is ignored. pad_type : str The type of padding to use. Passed directly to `torch.nn.functional.pad`, see PyTorch documentation for available options. stride : int The number of units to move each time convolution is applied. Passed to conv1d. Has no effect if `use_fft` is True. groups : int This option is passed to `conv1d` to split the input into groups for convolution. Input channels should be divisible by the number of groups. use_fft : bool When `use_fft` is passed `True`, then compute the convolution in the spectral domain using complex multiply. This is more efficient on CPU when the size of the kernel is large (e.g. reverberation). WARNING: Without padding, circular convolution occurs. This makes little difference in the case of reverberation, but may make more difference with different kernels. rotation_index : int This option only applies if `use_fft` is true. If so, the kernel is rolled by this amount before convolution to shift the output location. Returns ------- The convolved waveform. Example ------- >>> from speechbrain.dataio.dataio import read_audio >>> signal = read_audio('tests/samples/single-mic/example1.wav') >>> signal = signal.unsqueeze(0).unsqueeze(2) >>> kernel = torch.rand(1, 10, 1) >>> signal = convolve1d(signal, kernel, padding=(9, 0)) r z)Convolve1D expects a 3-dimensional tensorrr)r padmoder.Ndevice)r z1.6.0)n) signal_sizes)r weightstridegroupspadding)rr ValueError transposertuplernn functionalr7sizezerosr;catrparse __version__ torch.fftfftrfftirfftunbindstackconv1d)waveformkernelrApad_typer?r@use_fftrotation_index zero_lengthrH after_index before_indexrMresult convolvedf_signalf_kernelsig_realsig_imagker_realker_imagf_resultr&r&r' convolve1dsX7       rdcCs|j}t|jdkst|jdkrtt|jdkr#|dd}n t|jdkr/|d}t|jdkr?|dd}n t|jdkrK|d}t||d|}|jddd\}}t||d|d}t ||d||}t|dkr| d d}t|dkr| d}|S) a General function to contaminate a given signal with reverberation given a Room Impulse Response (RIR). It performs convolution between RIR and signal, but without changing the original amplitude of the signal. Arguments --------- waveforms : tensor The waveforms to normalize. Shape should be `[batch, time]` or `[batch, time, channels]`. rir_waveform : tensor RIR tensor, shape should be [time, channels]. rescale_amp : str Whether reverberated signal is rescaled (None) and with respect either to original signal "peak" amplitude or "avg" average amplitude. Choose between [None, "avg", "peak"]. Returns ------- waveforms: tensor Reverberated signal. r rrr9rT)axisr )rSrTrVrW) rrrrr(rGrrrdr5r*)r rir_waveform rescale_amp orig_shapeorig_amplitude value_max direct_indexr&r&r' reverberate?s:      rlcCs d|dS)a6Returns the amplitude ratio, converted from decibels. Arguments --------- SNR : float The ratio in decibels to convert. Returns ------- The amplitude ratio Example ------- >>> round(dB_to_amplitude(SNR=10), 3) 3.162 >>> dB_to_amplitude(SNR=0) 1.0 rr&)SNRr&r&r'r3s r3e皙?csd|kr dksJJ|ddksJ|dt|}||7}fdd}|d|||}|t|9}|t|}|d|||}|t|9}|t| }|d7<||dddS)aReturns a notch filter constructed from a high-pass and low-pass filter. (from https://tomroelandts.com/articles/ how-to-create-simple-band-pass-and-band-reject-filters) Arguments --------- notch_freq : float frequency to put notch as a fraction of the sampling rate / 2. The range of possible inputs is 0 to 1. filter_width : int Filter width in samples. Longer filters have smaller transition bands, but are more inefficient. notch_width : float Width of the notch, as a fraction of the sampling_rate / 2. Returns ------- The computed filter Example ------- >>> from speechbrain.dataio.dataio import read_audio >>> signal = read_audio('tests/samples/single-mic/example1.wav') >>> signal = signal.unsqueeze(0).unsqueeze(2) >>> kernel = notch_filter(0.25) >>> notched_signal = convolve1d(signal, kernel) rrrcs:dd}t||dtd||ddgS)zComputes the sinc function.cSst||S)N)rsin)xr&r&r'_sincsz)notch_filter..sinc.._sincNr)rrIones)rrrsr7r&r'sincs2znotch_filter..sincr r9)rarangeblackman_windowrview) notch_freq filter_width notch_widthinputsrvhlpfhhpfr&rur' notch_filters  rc Cs|dd}|dd\}}t||}||}||}||d|}||} |jg|d|R} td| d||} |  |j j } | d} |j g|| |R} | d| | | jg|dR} | S)aTaken from https://github.com/kaituoxu/Conv-TasNet/blob/master/src/utils.py Reconstructs a signal from a framed representation. Adds potentially overlapping frames of a signal with shape `[..., frames, frame_length]`, offsetting subsequent frames by `frame_step`. The resulting tensor has shape `[..., output_size]` where output_size = (frames - 1) * frame_step + frame_length Arguments --------- signal: A [..., frames, frame_length] torch.Tensor. All dimensions may be unknown, and rank must be at least 2. frame_step: int An integer denoting overlap offsets. Must be less than or equal to frame_length. Returns ------- A Tensor with shape [..., output_size] containing the overlap-added frames of signal's inner-most two dimensions. output_size = (frames - 1) * frame_step + frame_length Based on https://github.com/tensorflow/tensorflow/blob/r1.12/tensorflow/contrib/signal/python/ops/reconstruction_ops.py Example ------- >>> signal = torch.randn(5, 20) >>> overlapped = overlap_and_add(signal, 20) >>> overlapped.shape torch.Size([100]) Nrr9r)rGmathgcdryrrwunfoldclonedetachtor;type contiguous new_zeros index_add_) signal frame_stepouter_dimensionsframes frame_lengthsubframe_length subframe_stepsubframes_per_frame output_sizeoutput_subframessubframe_signalframer[r&r&r'overlap_and_adds2 rTc Cs||}t|dddddddf|dddddddf}tt|dttt|dtt|dfd}|||jdd}|rQt|dd}|S)aFunction for resynthesizing waveforms from enhanced mags. Arguments --------- enhanced_mag : torch.Tensor Predicted spectral magnitude, should be three dimensional. noisy_inputs : torch.Tensor The noisy waveforms before any processing, to extract phase. stft : torch.nn.Module Module for computing the STFT for extracting phase. istft : torch.nn.Module Module for computing the iSTFT for resynthesis. normalize_wavs : bool Whether to normalize the output wavs before returning them. Returns ------- enhanced_wav : torch.Tensor The resynthesized waveforms of the enhanced magnitudes with noisy phase. Nrrr9) sig_lengthr)r") ratan2mulrrIcosrqrr.) enhanced_mag noisy_inputsstftistftnormalize_wavs noisy_feats noisy_phasecomplex_predictions pred_wavsr&r&r' resynthesizes<   rcCsdttdtj|}ttjdd|dd|d ddd}|tj }|tj }tt tdtdtj|d|ddd}|tj d}|tj }|||S)a* Function for generating gabor impulse responses as used by GaborConv1d proposed in Neil Zeghidour, Olivier Teboul, F{'e}lix de Chaumont Quitry & Marco Tagliasacchi, "LEAF: A LEARNABLE FRONTEND FOR AUDIO CLASSIFICATION", in Proc of ICLR 2021 (https://arxiv.org/abs/2101.08596) ?@rrrr1g) rrtensorrpiexp tensordotrr complex64complex)tcenterfwhm denominatorgaussiancenter_frequency_complex t_complexsinusoidr&r&r'gabor_impulse_responseEs,    rc Csdttdtj|}ttjdd|dd|d ddd}tj|d|ddd}tj|j dd|j i}|d d d d dfd |d d d d df9<|d d d d f|d d d d df<tj ||j d }t|d d d d dft |d d d d df|d d d d df<t|d d d d dft |d d d d df|d d d d df<tj|j dd|j i}|d d|d d d d dft |d d|d d d d df|d d d d df<|d d|d d d d dft |d d|d d d d df|d d d d df<tj|j dd|j i} |d d d d df||d d d d dft || d d d d df<|d d d d dft ||d d d d df|| d d d d df<| S) aM Function for generating gabor impulse responses, but without using complex64 dtype as used by GaborConv1d proposed in Neil Zeghidour, Olivier Teboul, F{'e}lix de Chaumont Quitry & Marco Tagliasacchi, "LEAF: A LEARNABLE FRONTEND FOR AUDIO CLASSIFICATION", in Proc of ICLR 2021 (https://arxiv.org/abs/2101.08596) rrrrrr)rr;Nr9r:)rrrrrrrrrHrr; zeros_likerrqry) rrrrrtemptemp2rdenominator_sinusoidoutputr&r&r'%gabor_impulse_response_legacy_complexds<4&JJ & &r)Nrr)Nrr))rr/)rr)rr6rrFr)r)rorp)T)__doc__rr packagingrr(r.r2r5rdrlr3rrrrrr&r&r&r's,    D $ 4 {I ? ;-