在 librosa 中重新调整音频时面临问题

Question

我正在尝试使用我的数据集微调 wev2vec2 模型。因此我加载了音频。现在想将它们下采样到 16khz。但 librosa.reshape 函数给出了我无法解决的错误。错误是“resample() 需要 1 个位置参数，但给出了 3 个”

首先我尝试使用 sr 16khz 的 librosa 加载它。但由于我在这方面的经验较少。因此，我在项目的后期部分遇到了问题。我找到了一个以不同方式完成的代码。所以。我尝试使用他的方法，但现在面临问题。 这部分工作正常

database={}
audios = []
psr = []
for path in df['audio']:
  speech_array,sr = torchaudio.load(path)
  audios.append(speech_array[0].numpy())
  psr.append(sr)
database['audio'] = audios
database['psr'] = psr

每个索引都出现错误

import librosa
import numpy as np

# Assuming 'database' is your DataFrame containing 'audio' and 'psr' columns

# List to store new sampling rates
new_sr = []

# Resample each audio signal and store the new sampling rate
for i in range(len(database['psr'])):
    try:
        audio_signal = np.asarray(database['audio'][i])  # Convert audio to numpy array
        original_sr = database['psr'][i]  # Original sampling rate

        # Check if the audio signal is mono (single-channel)
        if audio_signal.ndim == 1:
            # Resample mono audio signal
            resampled_audio = librosa.resample(audio_signal, original_sr, 16000)
        else:
            # Resample each channel separately for multi-channel audio
            resampled_channels = []
            for channel in audio_signal:
                resampled_channel = librosa.resample(channel, original_sr, 16000)
                resampled_channels.append(resampled_channel)
            resampled_audio = np.array(resampled_channels)

        # Store resampled audio back in DataFrame
        database['audio'][i] = resampled_audio

        # Store new sampling rate (16000 Hz)
        new_sr.append(16000)
    except Exception as e:
        print(f"Error processing audio at index {i}: {e}")

# Add new sampling rates to the DataFrame
database['newsr'] = new_sr

Answer 1

这是

reshape

^[src]的定义：

@cache(level=20)
def resample(
    y: np.ndarray,
    *,  # forces you to pass all the following arguments only as named ones
    orig_sr: float,
    target_sr: float,
    res_type: str = "soxr_hq",
    fix: bool = True,
    scale: bool = False,
    axis: int = -1,
    **kwargs: Any,
) -> np.ndarray:

文档还提供了这样做的示例：

y, sr = librosa.load(librosa.ex('trumpet'), sr=22050)
y_8k = librosa.resample(y, orig_sr=sr, target_sr=8000)