o ij@sddlmZddlZddlZddlZddlZddlZddl Z ddl m Z ddZ ddZ dd Zd d Ze d d dddZddZddZddZddZdddZGdddZdS))print_functionN) lru_cachecCs$tj|sdStj|gdddS)zload segments file as arrayN))uttobject)recr)stf)etr)dtypendmin)ospathexistsnploadtxt) segments_filerW/home/ubuntu/.local/lib/python3.10/site-packages/funasr/models/eend/utils/kaldi_data.py load_segmentss rcCsPi}tj|s dSt|D]}|\}}}}|t|t|f||<q|SN)r rropenstripsplitfloatrretlinerrrr rrrload_segments_hashs  rcCshi}tj|s dSt|D]#}|\}}}}||vr"g||<|||t|t|dq|S)N)rrr )r rrrrrappendrrrrrload_segments_rechash&s   r cC ddt|D}dd|DS)z)return dictionary { rec: wav_rxfilename }cSg|] }|ddqSNr rr.0rrrr 4z load_wav_scp..cSi|] }|d|dqSrr rr&xrrr 5z load_wav_scp..r) wav_scp_filelinesrrr load_wav_scp2r2r )maxsizecCs|dr+tj|dddtjd}tjt|j dd\}}|||}||fS|dkrCtjt j dd\}}|||}||fStj|||d \}}||fS) a This function reads audio file and return data in numpy.float32 array. "lru_cache" holds recently loaded audio so that can be called many times on the same audio file. OPTIMIZE: controls lru_cache size for random access, considering memory size |NT)shellstdoutfloat32)r -)startstop) endswith subprocessPopenPIPEsfloadioBytesIOr8readsysstdin)wav_rxfilenamer;endpdata sampleraterrrload_wav8s    rMcCr!)z#returns dictionary { uttid: spkid }cSr"r#r$r%rrrr'Sr(z load_utt2spk..cSr)r*rr+rrrr-Tr.z load_utt2spk..r/) utt2spk_filer1rrr load_utt2spkQr3rOcC0tj|sdSddt|D}dd|DS)z,returns dictionary { spkid: list of uttids }NcSsg|]}|qSrr$r%rrrr'[sz load_spk2utt..cSsi|] }|d|ddqS)rr Nrr+rrrr-\z load_spk2utt..r rrr) spk2utt_filer1rrr load_spk2uttW rTcCrP)z&returns dictionary { recid: duration }NcSr"r#r$r%rrrr'cr(z!load_reco2dur..cSsi|] }|dt|dqSr*)rr+rrrr-drQz!load_reco2dur..rR) reco2dur_filer1rrr load_reco2dur_rUrWcCs"|dr ||dSd||S)zThis function returns preprocessed wav_rxfilename Args: wav_rxfilename: input process: command which can be connected via pipe, use stdin and stdout Returns: wav_rxfilename: output piped command r5z cat {} | {} |)r=format)rHprocessrrr process_wavgs  rZc cs|dur;|D]1}||d}t|\}}t|d|t}t|d|t}|d|||fVqdS|D]}t||\}}||fVq=dS)zThis function returns generator of segmented audio as (utterance id, numpy.float32 array) TODO?: sampling rate is not converted. Nrrr r)rMrrintastypeint) wavssegmentssegwavrKrL st_sample et_samplerrrrextract_segmentsxs    rdc@seZdZddZdddZdS) KaldiDatacCsx||_ttj|jd|_ttj|jd|_ttj|jd|_ t tj|jd|_ t tj|jd|_ dS)Nr_utt2spkzwav.scpreco2durspk2utt)data_dirr r rjoinr_rOrfr2r^rWrgrTrh)selfrirrr__init__s zKaldiData.__init__rNcCst|j|||\}}||fSr)rMr^)rkrecidr;rIrKraterrrrMszKaldiData.load_wavrN)__name__ __module__ __qualname__rlrMrrrrresreror) __future__rr rFnumpyrr>librosarArC functoolsrrrr r2rMrOrTrWrZrdrerrrrs(