一个Python函数实现Wav2Vec2中文语音转文字

可以使用下面的Python函数实现Wav2Vec2中文语音转文字：

import torch
import numpy as np
import pyaudio
from transformers import Wav2Vec2ForCTC, Wav2Vec2Tokenizer

def speech_to_text():
    # 加载模型和分词器
    model_name = "jonatasgrosman/wav2vec2-large-xlsr-53-chinese-zh-cn"
    tokenizer = Wav2Vec2Tokenizer.from_pretrained(model_name)
    model = Wav2Vec2ForCTC.from_pretrained(model_name)

    # 录音参数配置
    CHUNK = 1024
    FORMAT = pyaudio.paInt16
    CHANNELS = 1
    RATE = 16000
    RECORD_SECONDS = 5

    # 录音并预处理
    p = pyaudio.PyAudio()
    stream = p.open(format=FORMAT,
                    channels=CHANNELS,
                    rate=RATE,
                    input=True,
                    frames_per_buffer=CHUNK)

    frames = []
    print("开始录音，请讲话...")
    for i in range(0, int(RATE / CHUNK * RECORD_SECONDS)):
        data = stream.read(CHUNK)
        frames.append(data)

    print("录音结束！")
    stream.stop_stream()
    stream.close()
    p.terminate()

    audio = np.frombuffer(b''.join(frames), dtype=np.int16)
    input_values = tokenizer(audio, return_tensors="pt").input_values

    # 模型推理并转换成文本
    with torch.no_grad():
        logits = model(input_values).logits

    predicted_ids = torch.argmax(logits, dim=-1)
    transcription = tokenizer.batch_decode(predicted_ids)[0]

    return transcription

调用该函数即可进行中文语音转文字，例如：

text = speech_to_text()
print("语音转文字结果：", text)

需要注意的是，该函数使用的是Wav2Vec2模型，因此需要满足Wav2Vec2模型对输入语音信号的要求。如果输入的语音信号不符合要求，可能会影响模型的识别效果。