pydub音訊處理: 分離左右聲道並轉換成Numpy Array結構，以進行whisper語音辨識

decode_audio.py

import io
from typing import BinaryIO, Tuple, Union

import numpy as np
from pydub.utils import get_array_type

def decode_audio(
    input_file: Union[str, BinaryIO],
    sampling_rate: int = 16000,
) -> Tuple[np.ndarray, np.ndarray]:
    """Decodes the audio.

    Args:
      input_file: Path to the input file or a file-like object.
      sampling_rate: Resample the audio to this sample rate.

    Returns:
      A float32 Numpy array.

      returns a 2-tuple with the separated left and right channels.
    """
    raw_audio = AudioSegment.from_file(input_file)

    # 16-bit (2 bytes)
    raw_audio = raw_audio.set_sample_width(2)

    # 預設轉為雙聲道layout
    raw_audio = raw_audio.set_channels(2)

    # resampling
    raw_audio = raw_audio.set_frame_rate(sampling_rate)

    raw_data = raw_audio.raw_data

    dtype = get_array_type(raw_audio.sample_width * 8)
    audio = np.frombuffer(raw_data, dtype=dtype)

    # Convert s16 back to f32.
    audio = audio.astype(np.float32) / 32768.0
        
    left_channel = audio[0::2]
    right_channel = audio[1::2]
    return left_channel, right_channel