o Vi @sdZddlZddlZddlZddlZddlmZ ddl m Z m Z m Z mZddlmZddlmZd4ddZd d Zd5d d Zd5ddZd4ddZddZd6dededefddZdefddZdedejfddZdejdefd d!Z dede!fd"d#Z"d7d$ed%e ede ejeffd&d'Z#d(ejd)ed*eddfd+d,Z$ d7d-ejd.ejd*ed/e eddf d0d1Z%d-ejd.ejde eeffd2d3Z&dS)8z: Utility functions for audio processing and watermarking. N)TupleOptionalDictAny)sqrt)modeTc Csl|jd}g}td||D]&}||||}|r.|jd|kr.t|t||jd}||q |S)Nr)shaperangenpappendzeros)data window_lengthpad n_samplesframesidxchunkr?/home/ubuntu/.local/lib/python3.10/site-packages/perth/utils.py_signal_to_framess  rcCs t|SN)r hstack)rrrr_frames_to_signals rcCsL|jjdks Jd|d}d| d|d}}||||tjS)Nfz)This function takes floating point arrays)dtypekindclipastyper int16)wav bit_depthunsigned_bit_depth range_min range_maxrrr audio_to_rawsr)cCsT|d}d| d|d}}d\}}|||||||ddtjS)Nrr)gg?r)r!r"r float32)r$r%r&r'r(abrrrraw_pcm16_tofloat$s,r-cCs|t|dksJt|}t||r|dt|| }|r8t||kr8tt|t||j}t|}|d|S)Nr)lenr arraysplitTflatten_watermark)watermark_listlengthwrap watermarkrrrformatted_watermark+s   r7cCst|ddjdS)Nr)axisr)rsqueeze)watermark_vectorrrrr28r2@nkreturncCsH|dkrdSdtd|}dtd|}t||||tdS)Nrr)rint)r=r>alphabetarrrmodified_binets_fibonnaci<s rDr4cCs$tj|f}t||kddS)Nrr)r randomwheremean)r4r6rrrgenerate_dummy_watermarkHsrHr6cCstdd|DS)NcSsg|]}t|qSr)rA.0charrrr Nz*watermark_str_to_numpy..)r r/r6rrrwatermark_str_to_numpyMr;rOcCsddd|DS)Ncss|]}t|VqdSr)strrIrrr Rsz)watermark_numpy_to_str..)joinrNrrrwatermark_numpy_to_strQr;rTcCstdd|DS)NcSsg|]}|dvqS))10rrIrrrrLVrMz-validate_string_watermark..)anyrNrrrvalidate_string_watermarkUsrX audio_pathsrc CsJztj||d\}}||fWSty$}z td|d|d}~ww)z Load an audio file using librosa. Args: audio_path: Path to the audio file sr: Target sample rate. If None, the native sample rate is used. Returns: Tuple of (audio_data, sample_rate) )rZzCould not load audio file z: N)librosaload ExceptionIOError)rYrZaudio sample_rateerrr load_audioYs  rb audio_data file_pathr`cCs4tjtj|}tj|ddt|||dS)z Save audio data to a file. Args: audio_data: Audio data as a numpy array file_path: Output file path sample_rate: Sample rate for the audio file T)exist_okN)ospathdirnameabspathmakedirssfwrite)rcrdr` directoryrrr save_audioks rnoriginal watermarked output_pathc Csvtjdddd\}}tt||}|dj||ddd|dj||dd d|dd |dd |dd |d ||}|d|||dd |dd |ddt j t t |tjd}t jj|dd||dd|dd|dd |dd|j|djd|dddt|rt|tdStdS)a$ Plot a comparison between original and watermarked audio. Args: original: Original audio data watermarked: Watermarked audio data sample_rate: Sample rate of the audio output_path: Path to save the plot. If None, plot is shown interactively. r) )figsizergffffff?Original)rBlabel WatermarkedzWaveform ComparisonzTime (s) Amplitudez#Difference (Watermarked - Original) Difference)reftimelogr)x_axisy_axisrZaxzSpectrogram of DifferencezFrequency (Hz)z %+2.0f dB)rformatN)pltsubplotsr aranger.plot set_title set_xlabel set_ylabellegendr[amplitude_to_dbabsstftmaxdisplayspecshowcolorbar collections tight_layoutsavefigcloseshow) rorpr`rqfigaxsr|diffDrrrplot_audio_comparisonys4     rcCst|t|kr tdt||d}t|d}|}|dkr+dt||ntd}ttt|tt|}|dkrPdt|t|ntd}|||dS)ak Calculate audio quality metrics between original and watermarked audio. Args: original: Original audio data watermarked: Watermarked audio data Returns: Dictionary of quality metrics: - snr: Signal-to-Noise Ratio (dB) - mse: Mean Squared Error - psnr: Peak Signal-to-Noise Ratio (dB) z8Original and watermarked audio must have the same lengthrrrsinf)snrmsepsnr) r. ValueErrorr rGlog10floatrrr)rorpr signal_power noise_powerr max_valuerrrrcalculate_audio_metricss""(r)T)r)r<r)'__doc__rfnumpyr r[ soundfilerkmatplotlib.pyplotpyplotrtypingrrrrmathr scipy.statsrrrr)r-r7r2rArrDrHrQndarrayrOrTboolrXrbrnrrrrrrs>         & &/