o p i@sddlmZddlZddlmZmZddlZddlZddl m Z er.ddl m Z ddlm Z dd Zd(ddZ   d)d*ddZ   d+d,d&d'ZdS)-) annotationsN) TYPE_CHECKINGBinaryIO) AudioInfo)Path)TensorcCsd}d|d}|S)NZ paddleaudiozxonly PCM16 WAV supported. if want support more other audio types, please manually installed (usually with `pip install zQ`). and use paddle.audio.backends.set_backend('soundfile') to set audio backend)packageZwarn_msgr r i/home/app/PaddleOCR-VL/.venv_paddleocr/lib/python3.10/site-packages/paddle/audio/backends/wave_backend.py_error_message s r filepathstr | BinaryIOreturnrc Cst|dr|}nt|d}zt|}Wntjy,|d|t}t|w|}| }| }| d}d}|t |||||S)aGet signal information of input audio file. Args: filepath: audio path or file object. Returns: AudioInfo: info of the given audio. Example: .. code-block:: python >>> import os >>> import paddle >>> sample_rate = 16000 >>> wav_duration = 0.5 >>> num_channels = 1 >>> num_frames = sample_rate * wav_duration >>> wav_data = paddle.linspace(-1.0, 1.0, int(num_frames)) * 0.1 >>> waveform = wav_data.tile([num_channels, 1]) >>> base_dir = os.getcwd() >>> filepath = os.path.join(base_dir, "test.wav") >>> paddle.audio.save(filepath, waveform, sample_rate) >>> wav_info = paddle.audio.info(filepath) readrbrZPCM_S) hasattropenwaveErrorseekcloser NotImplementedError getnchannels getframerate getnframesZ getsampwidthr) r file_objfile_err_msgchannels sample_rateZ sample_framesbits_per_sampleencodingr r r info+s(     r$T str | Path frame_offsetint num_frames normalizeboolchannels_firsttuple[Tensor, int]cCs t|dr|}nt|d}zt|}Wntjy,|d|t}t|w|}| } | } | | } |t j | t jd} | t j} |rW| d}n| }t || |f}|dkrq||||ddf}t|}|rtj|ddgd }|| fS) aLoad audio data from file. load the audio content start form frame_offset, and get num_frames. Args: frame_offset: from 0 to total frames, num_frames: from -1 (means total frames) or number frames which want to read, normalize: if True: return audio which norm to (-1, 1), dtype=float32 if False: return audio with raw data, dtype=int16 channels_first: if True: return audio with shape (channels, time) Return: Tuple[paddle.Tensor, int]: (audio_content, sample rate) Examples: .. code-block:: python >>> import os >>> import paddle >>> sample_rate = 16000 >>> wav_duration = 0.5 >>> num_channels = 1 >>> num_frames = sample_rate * wav_duration >>> wav_data = paddle.linspace(-1.0, 1.0, int(num_frames)) * 0.1 >>> waveform = wav_data.tile([num_channels, 1]) >>> base_dir = os.getcwd() >>> filepath = os.path.join(base_dir, "test.wav") >>> paddle.audio.save(filepath, waveform, sample_rate) >>> wav_data_read, sr = paddle.audio.load(filepath) rrr)dtyper%Nr)perm)rrrrrrr rrrrZ readframesnpZ frombufferZint16astypefloat32ZreshapepaddleZ to_tensor transpose)r r'r)r*r,rrrr r!framesZ audio_contentZ audio_as_np16Z audio_as_np32Z audio_normZwaveformr r r load_s8 (      r7strsrcrr!r# str | Noner" int | NoneNonec Cs|jdks Jd|}|rt|}|jd}|dvr!tdt|d}|jtj kr4|d d}t |d } | || || || |Wd d S1s\wYd S) a Save audio tensor to file. Args: filepath: saved path src: the audio tensor sample_rate: the number of samples of audio per second. channels_first: src channel information if True, means input tensor is (channels, time) if False, means input tensor is (time, channels) encoding: audio encoding format, wave_backend only support PCM16 now. bits_per_sample: bits per sample, wave_backend only support 16 bits now. Returns: None Examples: .. code-block:: python >>> import paddle >>> sample_rate = 16000 >>> wav_duration = 0.5 >>> num_channels = 1 >>> num_frames = sample_rate * wav_duration >>> wav_data = paddle.linspace(-1.0, 1.0, int(num_frames)) * 0.1 >>> waveform = wav_data.tile([num_channels, 1]) >>> filepath = "./test.wav" >>> paddle.audio.save(filepath, waveform, sample_rate) zExpected 2D tensorr)Nr8z,Invalid bits_per_sample, only support 16 bitrr/zs(     6 S