o zi4@s>ddlZddlZddlZddlZddlmZGdddZdS)N)utilc@seZdZdZdZdZddZ d,dededefdd Z d,dededefd d Z defd d Z d-de j ejejefdefddZ d-de j ejejefdefddZ d.de j ejejefde j ejejefdefddZ d.de j ejejefde j ejejefdefddZ d.de j ejejefdefd d!Zd"e j ejejeffd#d$Zd%e j ejejeffd&d'Zd/d)efd*d+ZdS)0DSPMixinNcCsd|j|_|j|_t||j}t||j}||dkr$||}||}||||j|_||fS)Nr) batch_size_original_batch_size num_channels_original_num_channelsint sample_ratezero_pad signal_length_padded_signal_length)selfwindow_duration hop_duration window_length hop_lengthfactorrG/home/ubuntu/.local/lib/python3.10/site-packages/audiotools/core/dsp.py _preprocess_signal_for_windowings  z)DSPMixin._preprocess_signal_for_windowingTrr preprocessc cs|r |||\}}|jdd|j|_t|jD]&}d}||} ||}|d7}||} | |jkr5n ||d|| fVq#qdS)aFGenerator which yields windows of specified duration from signal with a specified hop length. Parameters ---------- window_duration : float Duration of every window in seconds. hop_duration : float Hop between windows in seconds. preprocess : bool, optional Whether to preprocess the signal, so that the first sample is in the middle of the first window, by default True Yields ------ AudioSignal Each window is returned as an AudioSignal. rrT.N)r audio_datareshaper ranger) rrrrrrbi start_idxend_idxrrrwindowss& zDSPMixin.windowscCsb|r |||\}}tjjj|jddd|jd|fd|fd}|ddddd|}||_|S)a Reshapes signal into windows of specified duration from signal with a specified hop length. Window are placed along the batch dimension. Use with :py:func:`audiotools.core.dsp.DSPMixin.overlap_and_add` to reconstruct the original signal. Parameters ---------- window_duration : float Duration of every window in seconds. hop_duration : float Hop between windows in seconds. preprocess : bool, optional Whether to preprocess the signal, so that the first sample is in the middle of the first window, by default True Returns ------- AudioSignal AudioSignal unfolded with shape ``(nb * nch * num_windows, 1, window_length)`` rr) kernel_sizestrider) rtorchnn functionalunfoldrrr permute)rrrrrrunfoldedrrrcollect_windowsFszDSPMixin.collect_windowsc Cst||j}|j}|j|j}}|j||d|ddd}tj j j |d|j fd|fd|fd}tj ||jd}tj j j |d|j fd|fd|fd}||}|||d}||_||||S)aFunction which takes a list of windows and overlap adds them into a signal the same length as ``audio_signal``. Parameters ---------- hop_duration : float How much to shift for each window (overlap is window_duration - hop_duration) in seconds. Returns ------- AudioSignal overlap-and-added signal. rrr#r) output_sizer!r"device)r r r rrrrr(r$r%r&foldr ones_liker-trim) rrrrnbnchr)foldednormrrroverlap_and_addns, zDSPMixin.overlap_and_add3cutoffszeroscCtt|d|j}||j}t|j}t|D]\}}tj | |d |j }||j|||<q||_d|_ |S)a9Low-passes the signal in-place. Each item in the batch can have a different low-pass cutoff, if the input to this signal is an array or tensor. If a float, all items are given the same low-pass filter. Parameters ---------- cutoffs : typing.Union[torch.Tensor, np.ndarray, float] Cutoff in Hz of low-pass filter. zeros : int, optional Number of taps to use in low-pass filter, by default 51 Returns ------- AudioSignal Low-passed AudioSignal. r#r8N)r ensure_tensorrr r$ empty_liker enumeratejulius LowPassFiltercputor- stft_data)rr7r8filteredrcutoff lp_filterrrrlow_pass  zDSPMixin.low_passcCr9)a?High-passes the signal in-place. Each item in the batch can have a different high-pass cutoff, if the input to this signal is an array or tensor. If a float, all items are given the same high-pass filter. Parameters ---------- cutoffs : typing.Union[torch.Tensor, np.ndarray, float] Cutoff in Hz of high-pass filter. zeros : int, optional Number of taps to use in high-pass filter, by default 51 Returns ------- AudioSignal High-passed AudioSignal. r#r:N)rr;rr r$r<rr=r>HighPassFilterr@rAr-rB)rr7r8rCrrD hp_filterrrr high_passrGzDSPMixin.high_passfmin_hzfmax_hzvalc Cs|j|j}}tj||jd}tj||jd}t||ks J|jd}tjd|j d||j d}|dddddf |j dd|jd}||k||k@}| |j }|||}|||}|td ||_|S) aMasks frequencies between ``fmin_hz`` and ``fmax_hz``, and fills them with the value specified by ``val``. Useful for implementing SpecAug. The min and max can be different for every item in the batch. Parameters ---------- fmin_hz : typing.Union[torch.Tensor, np.ndarray, float] Lower end of band to mask out. fmax_hz : typing.Union[torch.Tensor, np.ndarray, float] Upper end of band to mask out. val : float, optional Value to fill in, by default 0.0 Returns ------- AudioSignal Signal with ``stft_data`` manipulated. Apply ``.istft()`` to get the masked audio data. ndimrr#r,Nrr?) magnitudephaserr;rPr$allshapelinspacer r-repeatrrA masked_fillexprB) rrLrMrNmagrTnbinsbins_hzmaskrrrmask_frequenciess    zDSPMixin.mask_frequenciestmin_stmax_sc Cs|j|j}}tj||jd}tj||jd}t||ks J|jd}tjd|j ||j d}|dddddf |j d|jdd}||k||k@}| ||}| ||}|td||_|S) aMasks timesteps between ``tmin_s`` and ``tmax_s``, and fills them with the value specified by ``val``. Useful for implementing SpecAug. The min and max can be different for every item in the batch. Parameters ---------- tmin_s : typing.Union[torch.Tensor, np.ndarray, float] Lower end of timesteps to mask out. tmax_s : typing.Union[torch.Tensor, np.ndarray, float] Upper end of timesteps to mask out. val : float, optional Value to fill in, by default 0.0 Returns ------- AudioSignal Signal with ``stft_data`` manipulated. Apply ``.istft()`` to get the masked audio data. rOrrr,NrrQrR)rSrTrr;rPr$rUrVrWsignal_durationr-rXrrYrZrB) rr`rarNr[rTntbins_tr^rrrmask_timestepss   zDSPMixin.mask_timesteps db_cutoffcCs<|j}|}tj||jd}||k}|||}||_|S)a,Mask away magnitudes below a specified threshold, which can be different for every item in the batch. Parameters ---------- db_cutoff : typing.Union[torch.Tensor, np.ndarray, float] Decibel value for which things below it will be masked away. val : float, optional Value to fill in for masked portions, by default 0.0 Returns ------- AudioSignal Signal with ``stft_data`` manipulated. Apply ``.istft()`` to get the masked audio data. rO)rS log_magnituderr;rPrY)rrfrNr[log_magr^rrrmask_low_magnitudes3s zDSPMixin.mask_low_magnitudesshiftcCs"tj||jjd}|j||_|S)a`Shifts the phase by a constant value. Parameters ---------- shift : typing.Union[torch.Tensor, np.ndarray, float] What to shift the phase by. Returns ------- AudioSignal Signal with ``stft_data`` manipulated. Apply ``.istft()`` to get the masked audio data. rO)rr;rTrP)rrjrrr shift_phasePs zDSPMixin.shift_phasescalecCs.tj||jjd}|j|t|j|_|S)aCorrupts the phase randomly by some scaled value. Parameters ---------- scale : typing.Union[torch.Tensor, np.ndarray, float] Standard deviation of noise to add to the phase. Returns ------- AudioSignal Signal with ``stft_data`` manipulated. Apply ``.istft()`` to get the masked audio data. rO)rr;rTrPr$ randn_like)rrlrrr corrupt_phasebszDSPMixin.corrupt_phase333333?coefcCs^td| dgddd|j}|jdd|j}tjj j ||dd}|j|jj |_|S)a:Applies pre-emphasis to audio signal. Parameters ---------- coef : float, optional How much pre-emphasis to apply, lower values do less. 0 does nothing. by default 0.85 Returns ------- AudioSignal Pre-emphasized signal. rrr)padding) r$tensorviewrAr-rrr r%r&conv1drV)rrpkernelxrrr preemphasists $zDSPMixin.preemphasis)T)r6)rK)ro)__name__ __module__ __qualname__rrr rfloatboolr r*r5typingUnionr$Tensornpndarrayr rFrJr_rerirkrnrwrrrrr sv ( (, ! $ 1 . r)r}r>numpyrr$rrrrrrs