o si @s(dZddlZddlZddlmZddlmZddl Z ddl Z ddl Z ddl m Z dZdd Zd d Ze jd d dd+ddZe jd d d   d,ddZe jd d dd+ddZe jd d d   d,ddZddZ d-ddZdd Zd.d!d"Zd#d$Zd%d&Zd'd(Ze jd d dd)d*ZdS)/u .. warning:: The mir_eval.separation module is deprecated in mir_eval version 0.8, and will be removed. We recommend that you migrate your code to use an alternative package such as sigsep-museval https://sigsep.github.io/sigsep-mus-eval/ Source separation algorithms attempt to extract recordings of individual sources from a recording of a mixture of sources. Evaluation methods for source separation compare the extracted sources from reference sources and attempt to measure the perceptual quality of the separation. See also the bss_eval MATLAB toolbox: http://bass-db.gforge.inria.fr/bss_eval/ Conventions ----------- An audio signal is expected to be in the format of a 1-dimensional array where the entries are the samples of the audio signal. When providing a group of estimated or reference sources, they should be provided in a 2-dimensional array, where the first dimension corresponds to the source number and the second corresponds to the samples. Metrics ------- * :func:`mir_eval.separation.bss_eval_sources`: Computes the bss_eval_sources metrics from bss_eval, which optionally optimally match the estimated sources to the reference sources and measure the distortion and artifacts present in the estimated sources as well as the interference between them. * :func:`mir_eval.separation.bss_eval_sources_framewise`: Computes the bss_eval_sources metrics on a frame-by-frame basis. * :func:`mir_eval.separation.bss_eval_images`: Computes the bss_eval_images metrics from bss_eval, which includes the metrics in :func:`mir_eval.separation.bss_eval_sources` plus the image to spatial distortion ratio. * :func:`mir_eval.separation.bss_eval_images_framewise`: Computes the bss_eval_images metrics on a frame-by-frame basis. References ---------- .. [#vincent2006performance] Emmanuel Vincent, Rémi Gribonval, and Cédric Févotte, "Performance measurement in blind audio source separation," IEEE Trans. on Audio, Speech and Language Processing, 14(4):1462-1469, 2006. N)toeplitz) fftconvolve)utildcCs|j|jkrtd|j|j|jdks|jdkr$td|j|j|jdkr/tdnt|r7td|jdkrBtdnt|rJtd|jdtksX|jdtkrgtd |jd|jdtd S) a@Check that the input data to a metric are valid, and throws helpful errors if not. Parameters ---------- reference_sources : np.ndarray, shape=(nsrc, nsampl) matrix containing true sources estimated_sources : np.ndarray, shape=(nsrc, nsampl) matrix containing estimated sources z}The shape of estimated sources and the true sources should match. reference_sources.shape = {}, estimated_sources.shape = {}ztThe number of dimensions is too high (must be less than 3). reference_sources.ndim = {}, estimated_sources.ndim = {}rzureference_sources is empty, should be of size (nsrc, nsample). sdr, sir, sar, and perm will all be empty np.ndarrayszAll the reference sources should be non-silent (not all-zeros), but at least one of the reference sources is all 0s, which introduces ambiguity to the evaluation. (Otherwise we can add infinitely many all-zero sources.)zuestimated_sources is empty, should be of size (nsrc, nsample). sdr, sir, sar, and perm will all be empty np.ndarrayszAll the estimated sources should be non-silent (not all-zeros), but at least one of the estimated sources is all 0s. Since we require each reference source to be non-silent, having a silent estimated source will result in an underdetermined system.a The supplied matrices should be of shape (nsrc, nsampl) but reference_sources.shape[0] = {} and estimated_sources.shape[0] = {} which is greater than mir_eval.separation.MAX_SOURCES = {}. To override this check, set mir_eval.separation.MAX_SOURCES to a larger value.N) shape ValueErrorformatndimsizewarningswarn_any_source_silent MAX_SOURCES)reference_sourcesestimated_sourcesrG/home/ubuntu/.local/lib/python3.10/site-packages/mir_eval/separation.pyvalidateCsF      rc Cs.ttjtj|ttd|jddkddS)zDReturn true if the parameter sources has any silent first dimensions)axisrr)npanyallsumtupleranger )sourcesrrrrs&rz0.8z0.9)versionversion_removedTcCs|jdkr|tjddf}|jdkr|tjddf}t|||jdks+|jdkr=tgtgtgtgfS|jd}|rt||f}t||f}t||f}t|D],}t|D]%}t ||||d\} } } } t | | | | \|||f<|||f<|||f<qcq]t t t t|} tt| }t|}t| D]\}}t|||f||<q| t|}||f}||||||t|fSt|}t|}t|}t|D]}t ||||d\} } } } t | | | | \||<||<||<qt|}||||fS)a: Ordering and measurement of the separation quality for estimated source signals in terms of filtered true source, interference and artifacts. The decomposition allows a time-invariant filter distortion of length 512, as described in Section III.B of [#vincent2006performance]_. Passing ``False`` for ``compute_permutation`` will improve the computation performance of the evaluation; however, it is not always appropriate and is not the way that the BSS_EVAL Matlab toolbox computes bss_eval_sources. Examples -------- >>> # reference_sources[n] should be an ndarray of samples of the >>> # n'th reference source >>> # estimated_sources[n] should be the same for the n'th estimated >>> # source >>> (sdr, sir, sar, ... perm) = mir_eval.separation.bss_eval_sources(reference_sources, ... estimated_sources) Parameters ---------- reference_sources : np.ndarray, shape=(nsrc, nsampl) matrix containing true sources (must have same shape as estimated_sources) estimated_sources : np.ndarray, shape=(nsrc, nsampl) matrix containing estimated sources (must have same shape as reference_sources) compute_permutation : bool, optional compute permutation of estimate/source combinations (True by default) Returns ------- sdr : np.ndarray, shape=(nsrc,) vector of Signal to Distortion Ratios (SDR) sir : np.ndarray, shape=(nsrc,) vector of Source to Interference Ratios (SIR) sar : np.ndarray, shape=(nsrc,) vector of Sources to Artifacts Ratios (SAR) perm : np.ndarray, shape=(nsrc,) vector containing the best ordering of estimated sources in the mean SIR sense (estimated source number ``perm[j]`` corresponds to true source number ``j``). Note: ``perm`` will be ``[0, 1, ..., nsrc-1]`` if ``compute_permutation`` is ``False``. References ---------- .. [#] Emmanuel Vincent, Shoko Araki, Fabian J. Theis, Guido Nolte, Pau Bofill, Hiroshi Sawada, Alexey Ozerov, B. Vikrham Gowreesunker, Dominik Lutter and Ngoc Q.K. Duong, "The Signal Separation Evaluation Campaign (2007-2010): Achievements and remaining challenges", Signal Processing, 92, pp. 1928-1936, 2012. rNr)r rnewaxisrr arrayremptyr_bss_decomp_mtifilt_bss_source_critlist itertools permutationslenarange enumeratemeanargmaxasarray)rrcompute_permutationnsrcsdrsirsarjestjtrues_truee_spate_interfe_artifpermsmean_sirdumipermpoptidxjrrrbss_eval_sourcessN :  $     $        "  rC/ FcCs|jdkr|tjddf}|jdkr|tjddf}t|||jdks+|jdkr=tgtgtgtgfS|jd}tt|jd|||}|dkrct |||}dd|DSt ||f}t ||f} t ||f} t ||f} t |D]i} t | || ||} |dd| f}|dd| f}t |st |st |||\|dd| f<| dd| f<| dd| f<| dd| f<qtj|dd| f<| dd| f<| dd| f<| dd| f<q|| | | fS)a Framewise computation of bss_eval_sources Please be aware that this function does not compute permutations (by default) on the possible relations between reference_sources and estimated_sources due to the dangers of a changing permutation. Therefore (by default), it assumes that ``reference_sources[i]`` corresponds to ``estimated_sources[i]``. To enable computing permutations please set ``compute_permutation`` to be ``True`` and check that the returned ``perm`` is identical for all windows. NOTE: if ``reference_sources`` and ``estimated_sources`` would be evaluated using only a single window or are shorter than the window length, the result of :func:`mir_eval.separation.bss_eval_sources` called on ``reference_sources`` and ``estimated_sources`` (with the ``compute_permutation`` parameter passed to :func:`mir_eval.separation.bss_eval_sources`) is returned. Examples -------- >>> # reference_sources[n] should be an ndarray of samples of the >>> # n'th reference source >>> # estimated_sources[n] should be the same for the n'th estimated >>> # source >>> (sdr, sir, sar, ... perm) = mir_eval.separation.bss_eval_sources_framewise( reference_sources, ... estimated_sources) Parameters ---------- reference_sources : np.ndarray, shape=(nsrc, nsampl) matrix containing true sources (must have the same shape as ``estimated_sources``) estimated_sources : np.ndarray, shape=(nsrc, nsampl) matrix containing estimated sources (must have the same shape as ``reference_sources``) window : int, optional Window length for framewise evaluation (default value is 30s at a sample rate of 44.1kHz) hop : int, optional Hop size for framewise evaluation (default value is 15s at a sample rate of 44.1kHz) compute_permutation : bool, optional compute permutation of estimate/source combinations for all windows (False by default) Returns ------- sdr : np.ndarray, shape=(nsrc, nframes) vector of Signal to Distortion Ratios (SDR) sir : np.ndarray, shape=(nsrc, nframes) vector of Source to Interference Ratios (SIR) sar : np.ndarray, shape=(nsrc, nframes) vector of Sources to Artifacts Ratios (SAR) perm : np.ndarray, shape=(nsrc, nframes) vector containing the best ordering of estimated sources in the mean SIR sense (estimated source number ``perm[j]`` corresponds to true source number ``j``). Note: ``perm`` will be ``range(nsrc)`` for all windows if ``compute_permutation`` is ``False`` rNrrcSg|]}t|dqSr expand_dims.0scorerrr [z.bss_eval_sources_framewise..)r rr"rr r#rintfloorrCr$rslicernan)rrwindowhopr0r1nwinresultr2r3r4r?k win_slice ref_slice est_slicerrrbss_eval_sources_framewises8 E  $   >D r\c Cst|}t|}t|||jdks|jdkr/tgtgtgtgtgfS|jd}|jd}|jd}|rt||f}t||f}t||f}t||f} t|D]9} t|D]2} t|tj || ||fdd| d\} } }}t | | ||\|| | f<|| | f<|| | f<| | | f<qfq`t t t t|}tt|}t|}t|D]\}}t|||f||<q|t|}||f}||||||| |t|fSt|}t|}t|}t|} dg|}td}t|D]5}t|tj ||||fdd|d|||\} } }}}}|||<t | | ||\||<||<||<| |<qt|}|||| |fS)a Compute the bss_eval_images function from the BSS_EVAL Matlab toolbox. Ordering and measurement of the separation quality for estimated source signals in terms of filtered true source, interference and artifacts. This method also provides the ISR measure. The decomposition allows a time-invariant filter distortion of length 512, as described in Section III.B of [#vincent2006performance]_. Passing ``False`` for ``compute_permutation`` will improve the computation performance of the evaluation; however, it is not always appropriate and is not the way that the BSS_EVAL Matlab toolbox computes bss_eval_images. Examples -------- >>> # reference_sources[n] should be an ndarray of samples of the >>> # n'th reference source >>> # estimated_sources[n] should be the same for the n'th estimated >>> # source >>> (sdr, isr, sir, sar, ... perm) = mir_eval.separation.bss_eval_images(reference_sources, ... estimated_sources) Parameters ---------- reference_sources : np.ndarray, shape=(nsrc, nsampl, nchan) matrix containing true sources estimated_sources : np.ndarray, shape=(nsrc, nsampl, nchan) matrix containing estimated sources compute_permutation : bool, optional compute permutation of estimate/source combinations (True by default) Returns ------- sdr : np.ndarray, shape=(nsrc,) vector of Signal to Distortion Ratios (SDR) isr : np.ndarray, shape=(nsrc,) vector of source Image to Spatial distortion Ratios (ISR) sir : np.ndarray, shape=(nsrc,) vector of Source to Interference Ratios (SIR) sar : np.ndarray, shape=(nsrc,) vector of Sources to Artifacts Ratios (SAR) perm : np.ndarray, shape=(nsrc,) vector containing the best ordering of estimated sources in the mean SIR sense (estimated source number ``perm[j]`` corresponds to true source number ``j``). Note: ``perm`` will be ``(1,2,...,nsrc)`` if ``compute_permutation`` is ``False``. References ---------- .. [#] Emmanuel Vincent, Shoko Araki, Fabian J. Theis, Guido Nolte, Pau Bofill, Hiroshi Sawada, Alexey Ozerov, B. Vikrham Gowreesunker, Dominik Lutter and Ngoc Q.K. Duong, "The Signal Separation Evaluation Campaign (2007-2010): Achievements and remaining challenges", Signal Processing, 92, pp. 1928-1936, 2012. rrrForderr!)r atleast_3drr r#rr$r_bss_decomp_mtifilt_imagesreshape_bss_image_critr'r(r)r*r+r,r-r.r/zeros)rrr0r1nsamplnchanr2isrr3r4r5r6r7r8r9r:r;r<r=r>r?r@rAGjGrBGj_temprrrbss_eval_imagestsr =  ,           $        rkc Cst|}t|}t|||jdks|jdkr+tgtgtgtgfS|jd}tt|jd|||}|dkrQt|||}dd|DSt ||f}t ||f} t ||f} t ||f} t ||f} t |D]v} t | || ||}|dd|ddf}|dd|ddf}t |st |st|||\|dd| f<| dd| f<| dd| f<| dd| f<| dd| f<qxtj |dd| f<| dd| f<| dd| f<| dd| f<qx|| | | | fS)aa Framewise computation of bss_eval_images Please be aware that this function does not compute permutations (by default) on the possible relations between ``reference_sources`` and ``estimated_sources`` due to the dangers of a changing permutation. Therefore (by default), it assumes that ``reference_sources[i]`` corresponds to ``estimated_sources[i]``. To enable computing permutations please set ``compute_permutation`` to be ``True`` and check that the returned ``perm`` is identical for all windows. NOTE: if ``reference_sources`` and ``estimated_sources`` would be evaluated using only a single window or are shorter than the window length, the result of ``bss_eval_images`` called on ``reference_sources`` and ``estimated_sources`` (with the ``compute_permutation`` parameter passed to ``bss_eval_images``) is returned Examples -------- >>> # reference_sources[n] should be an ndarray of samples of the >>> # n'th reference source >>> # estimated_sources[n] should be the same for the n'th estimated >>> # source >>> (sdr, isr, sir, sar, ... perm) = mir_eval.separation.bss_eval_images_framewise( reference_sources, ... estimated_sources, window, .... hop) Parameters ---------- reference_sources : np.ndarray, shape=(nsrc, nsampl, nchan) matrix containing true sources (must have the same shape as ``estimated_sources``) estimated_sources : np.ndarray, shape=(nsrc, nsampl, nchan) matrix containing estimated sources (must have the same shape as ``reference_sources``) window : int Window length for framewise evaluation hop : int Hop size for framewise evaluation compute_permutation : bool, optional compute permutation of estimate/source combinations for all windows (False by default) Returns ------- sdr : np.ndarray, shape=(nsrc, nframes) vector of Signal to Distortion Ratios (SDR) isr : np.ndarray, shape=(nsrc, nframes) vector of source Image to Spatial distortion Ratios (ISR) sir : np.ndarray, shape=(nsrc, nframes) vector of Source to Interference Ratios (SIR) sar : np.ndarray, shape=(nsrc, nframes) vector of Sources to Artifacts Ratios (SAR) perm : np.ndarray, shape=(nsrc, nframes) vector containing the best ordering of estimated sources in the mean SIR sense (estimated source number perm[j] corresponds to true source number j) Note: perm will be range(nsrc) for all windows if compute_permutation is False rrrcSrFrGrIrKrrrrNTrOz-bss_eval_images_framewise..N)rr`rr r#rrPrQrkr$rrRrrS)rrrTrUr0r1rVrWr2rgr3r4r?rXrYrZr[rrrbss_eval_images_framewises6 H  $   LDrlc Cs|j}t||t|df}t||tjddf|||}t|||||}| ||}|d||7<||||fS)aDecomposition of an estimated source image into four components representing respectively the true source image, spatial (or filtering) distortion, interference and artifacts, derived from the true source images using multichannel time-invariant filters. rN)r rhstackrd_projectr") restimated_sourcerBflenrer7r8r9r:rrrr%ns r%c Cs*t|d}t|d}|duo|du}ttj||||fddt||dff} |rDt||tjddf|||\} }nt||tjddf||} | | } |rbt||||\} }nt|||} | | | } | | | } | ddd|f|7<|r| | | | ||fS| | | | fS)axDecomposition of an estimated source image into four components representing respectively the true source image, spatial (or filtering) distortion, interference and artifacts, derived from the true source images using multichannel time-invariant filters. Adapted version to work with multichannel sources. Improved performance can be gained by passing Gj and G parameters initially as all zeros. These parameters store the results from the computation of the G matrix in _project_images and then return them for subsequent calls to this function. This only works when not computing permutations. rrNr]r^)rrrmrb transposerd_project_imagesr") rrorBrprhrirerfsavegr7r8r9r:rrrras2      rac Cs|jd}|jd}t|t||dff}t|t|df}tdtt||d}tjj ||dd}tjj ||d}t||||f}t |D]a} t |D]Z} || t || } t tj | } tt| d| d| df| d|d } | || || d|| || d|f<| j|| || d|| || d|f<qZqTt||} t |D]-} || t |}t tj |}t|d|d| df| | || d|<qztj|| j||d d }Wntjjjytj|| dj||d d }Ynwt||d}t |D]} |t|dd| f|| d||d7}q(|S) zLeast-squares projection of estimated source on the subspace spanned by delayed versions of reference sources, with delays between 0 and flen-1 rrr?nrrvrHNrr]r^)rrrmrdrPceillog2scipyfftpackfftrconjrealifftrTlinalgsolverb LinAlgErrorlstsqr)rrorpr1ren_fftsfsefrir>rBssfssDssefCsprojrrrrns<     .,0 4" 2rnc CsP|jd}|jd}|jd}tjt|d|||fdd}t|t|||dff}t|t||dff}tdtt||d}t j j ||dd}t j j ||d } |d urd } t||||||f}t ||D]c} t | dD]Z} || t || } tt j | } tt| d| d | d f| d |d }||| || d|| || d|f<|j|| || d|| || d|f<qq{nd} t|dkrdt||||||f}t ||D]d} t | dD][} || t || } tt j | } tt| d| d | d f| d |d }||| || d|| || d|f<|j|| || d|| || d|f<qqt||||f}t ||D]<}t |D]4} ||t | | }tt j |}t|d|d | d f||||d|| f<q{quztj||j||||dd}Wntjjjytj||dj||||dd}Ynwt|||df}t ||D])}t |D]!} || t|d d || f||d ||d7<qq| r&||fS|S)aYLeast-squares projection of estimated source on the subspace spanned by delayed versions of reference sources, with delays between 0 and flen-1. Passing G as all zeros will populate the G matrix and return it so it can be passed into the next call to avoid recomputing G (this will only works if not computing permutations). rrr)rrrr]r^rtrurwNFrHrxT)rrrbrqrmrdrPrzr{r|r}r~rrrrrrrrrrrr)rrorprir1rerfrrrrsr>rBrrrrXrrrrrrrrs    .,0.,2   $( & rrcCst||}tt|dt||d}tt|dt|d}tt||dt|d}|||fS)zMeasurement of the separation quality for a given source in terms of filtered true source, interference and artifacts. r_safe_dbrr)r7r8r9r:s_filtr2r3r4rrrr&3s "" r&cCstt|dt|||d}tt|dt|d}tt||dt|d}tt|||dt|d}||||fS)zMeasurement of the separation quality for a given image in terms of filtered true source, spatial error, interference and artifacts. rr)r7r8r9r:r2rgr3r4rrrrc?s &"& rccCs |dkrtjSdt||S)zProperly handle the potential +Inf db SIR, instead of raising a RuntimeWarning. Only denominator is checked because the numerator can never be 0. r )rinflog10)numdenrrrrKsrc Kstt}tjt||fi|\}}}}}||d<||d<||d<||d<||d<tjt||fi|\}}}}}||d<||d<||d<||d <||d <|jd kr|jd krtjt||fi|\}}}}||d <||d <||d<||d<tjt ||fi|\}}}}||d<||d<||d<||d<|S)aCompute all metrics for the given reference and estimated signals. NOTE: This will always compute :func:`mir_eval.separation.bss_eval_images` for any valid input and will additionally compute :func:`mir_eval.separation.bss_eval_sources` for valid input with fewer than 3 dimensions. Examples -------- >>> # reference_sources[n] should be an ndarray of samples of the >>> # n'th reference source >>> # estimated_sources[n] should be the same for the n'th estimated source >>> scores = mir_eval.separation.evaluate(reference_sources, ... estimated_sources) Parameters ---------- reference_sources : np.ndarray, shape=(nsrc, nsampl[, nchan]) matrix containing true sources estimated_sources : np.ndarray, shape=(nsrc, nsampl[, nchan]) matrix containing estimated sources **kwargs Additional keyword arguments which will be passed to the appropriate metric or preprocessing functions. Returns ------- scores : dict Dictionary of scores, where the key is the metric name (str) and the value is the (float) score achieved. zImages - Source to DistortionzImages - Image to SpatialzImages - Source to InterferencezImages - Source to ArtifactzImages - Source permutationz$Images Frames - Source to Distortionz Images Frames - Image to Spatialz&Images Frames - Source to Interferencez"Images Frames - Source to Artifactz"Images Frames - Source permutationrz%Sources Frames - Source to Distortionz'Sources Frames - Source to Interferencez#Sources Frames - Source to Artifactz#Sources Frames - Source permutationzSources - Source to Distortionz Sources - Source to InterferencezSources - Source to ArtifactzSources - Source permutation) collections OrderedDictr filter_kwargsrktolistrlr r\rC) rrkwargsscoresr2rgr3r4r?rrrevaluateUsR#                  r)T)rDrEF)NN)N)__doc__numpyr scipy.fftpackr| scipy.linalgr scipy.signalrrr(r rrrr deprecatedrCr\rkrlr%rarnrrr&rcrrrrrrsJ4   K   n o   r 3 ,P