o zi4)@shddlZddlZddlZddlZddlZddlmmZ ddl Z Gdddejj Z GdddZ dS)Nc seZdZdZ    ddededed ed ef fd d Zd e j fddZ d e j fddZ d e j fddZ d e j fddZddZd e j fddZeddZejddZZS)MeteraTensorized version of pyloudnorm.Meter. Works with batched audio tensors. Parameters ---------- rate : int Sample rate of audio. filter_class : str, optional Class of weighting filter used. K-weighting' (default), 'Fenton/Lee 1' 'Fenton/Lee 2', 'Dash et al.' by default "K-weighting" block_size : float, optional Gating block size in seconds, by default 0.400 zeros : int, optional Number of zeros to use in FIR approximation of IIR filters, by default 512 use_fir : bool, optional Whether to use FIR approximation or exact IIR formulation. If computing on GPU, ``use_fir=True`` will be used, as its much faster, by default False K-weighting皙?Frate filter_class block_sizezerosuse_firc st||_||_||_||_tt gd}| d|t |f}d|d<t t |j d|f}t t |j } t|j D]\} \} } tj| j| j||| <| j| | <qEt|ddddf}| d|| d | dS) N)?r r (\?r Gr ).r.firs passband_gain)super__init__rrrr torch from_numpynparrayregister_bufferr len_filters enumerateitemsscipysignallfilterbarcopyfloat) selfrrrr r r impulserri_ filter_stage __class__L/home/ubuntu/.local/lib/python3.10/site-packages/audiotools/core/loudness.pyr"s"      zMeter.__init__datacCs|j\}}}|ddd}|||d|}|jjd}t|jjdD](}t|||f}tj ||j|ddf}|j ||}|dd|df}q$|ddd}|ddd|ddf}|S)aPerforms FIR approximation of loudness computation. Parameters ---------- data : torch.Tensor Audio data of shape (nb, nch, nt). Returns ------- torch.Tensor Filtered audio data. rrrN.) shapepermutereshaperrangeFpadjuliusfftconv fft_conv1dr)r$r-nbntnch pad_lengthr&r+r+r,apply_filter_gpuEs  zMeter.apply_filter_gpuc Cs|jD]9\}}|j}t|j|j}t|j |j}| ddd}t j j |||dd}|| ddd}q|S)aPerforms IIR formulation of loudness computation. Parameters ---------- data : torch.Tensor Audio data of shape (nb, nch, nt). Returns ------- torch.Tensor Filtered audio data. rr.rF)clamp)rrrrrr!r#todevicer r0 torchaudio functionalr) r$r-r'r(ra_coeffsb_coeffs_datafilteredr+r+r,apply_filter_cpufs zMeter.apply_filter_cpucCs(|js|jr ||}|S||}|S)aXApplies filter on either CPU or GPU, depending on if the audio is on GPU or is on CPU, or if ``self.use_fir`` is True. Parameters ---------- data : torch.Tensor Audio data of shape (nb, nch, nt). Returns ------- torch.Tensor Filtered audio data. )is_cudar r<rFr$r-r+r+r, apply_filters   zMeter.apply_filtercCs ||S)Computes integrated loudness of data. Parameters ---------- data : torch.Tensor Audio data of shape (nb, nch, nt). Returns ------- torch.Tensor Filtered audio data. )integrated_loudnessrHr+r+r,forwards z Meter.forwardcCs\|j}d}d|}t||j}t||j|}tj|ddd||}|dd}|S)Ng?r rr.rr)rintrr5coreunfoldr0 transpose)r$ input_dataT_goverlapstep kernel_sizestrideunfoldedr+r+r,_unfolds z Meter._unfoldc Cst|s t|}n|}t|}|jdkr |d}|jdkr*|d}|j\}}}||}|j }|j }d}| |} d||j | d} ddt|d d |d f| jd d d } | | } | } d| | |k<| |k} | d| d} ddt| |d d |fdd}|d d d d f}|||| jd}| } d| | |k<d| | |k<| |k| |k} | d| d} t| t| | } tttjj| | td k<tttjj| | td  k<ddt|d d |f| d }|S)rJr.rrgQr g&1g$@NrT)keepdiminf)r is_tensorrr#r"ndim unsqueezer/rIr rrYrsquaresumlog10 expand_asexpandwhereisnan zeros_likerfinfofloat32maxmin)r$r-rRr8r9r:r rSGamma_arXzl z_avg_gatedmaskedGamma_rLUFSr+r+r,rKsJ         .  *   (zMeter.integrated_loudnesscCs|jS)N) _filter_class)r$r+r+r,rszMeter.filter_classcCs.ddlm}||j}||_||_|j|_dS)Nr)r) pyloudnormrrrrsr)r$valuermeterr+r+r,rs   )rrrF)__name__ __module__ __qualname____doc__rNstrr#boolrrTensorr<rFrIrLrYrKpropertyrsetter __classcell__r+r+r)r,r s6#! I rc@s,eZdZdZdZ d dedefddZdS) LoudnessMixinNirrrrc Ks|jdur |j|jS|j}|jdkr$td|j|j}|d|t|jf||d|}||j}| |j ddd}| |t j||jd|j}t |||_|j|jS)aCalculates loudness using an implementation of ITU-R BS.1770-4. Allows control over gating block size and frequency weighting filters for additional control. Measure the integrated gated loudness of a signal. API is derived from PyLoudnorm, but this implementation is ported to PyTorch and is tensorized across batches. When on GPU, an FIR approximation of the IIR filters is used to compute loudness for speed. Uses the weighting filters and block size defined by the meter the integrated loudness is measured based upon the gating algorithm defined in the ITU-R BS.1770-4 specification. Parameters ---------- filter_class : str, optional Class of weighting filter used. K-weighting' (default), 'Fenton/Lee 1' 'Fenton/Lee 2', 'Dash et al.' by default "K-weighting" block_size : float, optional Gating block size in seconds, by default 0.400 kwargs : dict, optional Keyword arguments to :py:func:`audiotools.core.loudness.Meter`. Returns ------- torch.Tensor Loudness of audio data. Ng?r)rrr.r)r?) _loudnessr>r? signal_lengthsignal_durationrN sample_ratezero_padrrK audio_datar0truncate_samplesr ones_like MIN_LOUDNESSmaximum) r$rrkwargsoriginal_lengthpad_lenrvloudness min_loudnessr+r+r,r s(     zLoudnessMixin.loudness)rr)rwrxryrrr{r#rr+r+r+r,rsr)r"r5numpyrrrtorch.nn.functionalnnrAr3r@Modulerrr+r+r+r,s}