使用librosa隔离音频前景并转换回音频流

import librosa

audio_path = 'path_to_audio_file.wav'
audio, sr = librosa.load(audio_path)

# 提取音频的短时傅里叶变换（STFT）特征
stft = librosa.stft(audio)

# 计算音频的能量谱
energy = librosa.feature.rms(S=stft)

# 根据能量谱选择前景音频
foreground_stft = stft * (energy >= energy.mean())

# 将前景音频转换回时域信号
foreground_audio = librosa.istft(foreground_stft)

output_path = 'path_to_output_file.wav'
librosa.output.write_wav(output_path, foreground_audio, sr)