使用pyannote.audio进行语音分离和说话人识别

pip install pyannote.audio

# _*_ coding: utf-8 _*_
# @Time : 2024/3/16 10:47
# @Author : Michael
# @File : spearker_rec.py
# @desc :
import torch
from pyannote.audio import Model, Pipeline, Inference
from pyannote.core import Segment
from scipy.spatial.distance import cosine


def extract_speaker_embedding(pipeline, audio_file, speaker_label):
    diarization = pipeline(audio_file)
    speaker_embedding = None
    for turn, _, label in diarization.itertracks(yield_label=True):
        if label == speaker_label:
            segment = Segment(turn.start, turn.end)
            speaker_embedding = inference.crop(audio_file, segment)
            break
    return speaker_embedding

# 对于给定的音频，提取声纹特征并与人库中的声纹进行比较
def recognize_speaker(pipeline, audio_file):
    diarization = pipeline(audio_file)
    speaker_turns = []
    for turn, _, speaker_label in diarization.itertracks(yield_label=True):
        # 提取切片的声纹特征
        embedding = inference.crop(audio_file, turn)  
        distances = {}
        for speaker, embeddings in speaker_embeddings.items():  
	        # 计算与已知说话人的声纹特征的余弦距离
            distances[speaker] = min([cosine(embedding, e) for e in embeddings])
        # 选择距离最小的说话人
        recognized_speaker = min(distances, key=distances.get)  
        speaker_turns.append((turn, recognized_speaker))  
        # 记录说话人的时间段和余弦距离最小的预测说话人
    return speaker_turns

if __name__ == "__main__":
    token = "hf_***"  # 请替换为您的Hugging Face Token

    # 加载声音分离识别模型
    pipeline = Pipeline.from_pretrained(
        "pyannote/speaker-diarization-3.1",
        use_auth_token=token,  # 在项目页面agree使用协议，并获取 Hugging Face Token
        # cache_dir="/home/huggingface/hub/models--pyannote--speaker-diarization-3.1/"
    )

    # 加载声纹嵌入模型
    embed_model = Model.from_pretrained("pyannote/embedding", use_auth_token=token)
    inference = Inference(embed_model, window="whole")

    # pipeline.to(torch.device("cuda"))

    # 假设您已经有一个包含不同人声的音频文件集,以及对应的人
    audio_files = {
        "mick": "mick.wav",  # mick的音频
        "moon": "moon.wav",  # moon的音频
    }
    speaker_embeddings = {}
    for speaker, audio_file in audio_files.items():
        diarization = pipeline(audio_file)
        for turn, _, speaker_label in diarization.itertracks(yield_label=True):
            embedding = extract_speaker_embedding(pipeline, audio_file, speaker_label)
            # 获取原始已知说话人的声纹特征
            speaker_embeddings.setdefault(speaker, []).append(embedding)

    # 给定新的未知人物的音频文件
    given_audio_file = "2_voice.wav"  # 前半部分是 mick 说话，后半部分是 moon 说话

    # 识别给定音频中的说话人
    recognized_speakers = recognize_speaker(pipeline, given_audio_file)
    print("Recognized speakers in the given audio:")
    for turn, speaker in recognized_speakers:
        print(f"Speaker {speaker} spoke between {turn.start:.2f}s and {turn.end:.2f}s")

Model was trained with pyannote.audio 0.0.1, yours is 3.1.1. Bad things might happen unless you revert pyannote.audio to 0.x.
Model was trained with torch 1.8.1+cu102, yours is 2.2.1+cpu. Bad things might happen unless you revert torch to 1.x.

Recognized speakers in the given audio:
Speaker mick spoke between 0.57s and 1.67s
Speaker moon spoke between 2.47s and 2.81s
Speaker moon spoke between 3.08s and 4.47s