1. 环境

python版本：3.11.9

2.完整代码

import sqlite3
import time

from funasr import AutoModel
import sounddevice as sd
import numpy as np
from pypinyin import lazy_pinyin

# 模型参数设置
chunk_size = [0, 10, 5]
encoder_chunk_look_back = 7
decoder_chunk_look_back = 5

model = AutoModel(model="D:\SpeechRecognize\speech_paraformer-large-vad-punc_asr_nat-zh-cn-16k-common-vocab8404-pytorch")

# 假设模型要求的采样率为 16000
fs = 16000
duration = 3 #时间
chunk_stride = chunk_size[1] * 960
cache = {}
window_size = 3

# 连接到 SQLite 数据库，如果不存在则会创建新的数据库文件
conn = sqlite3.connect('speech_recognition.db')
cursor = conn.cursor()

# 创建表格
cursor.execute('''
    CREATE TABLE IF NOT EXISTS speech_data
    (text TEXT, time_stamp TEXT, batch TEXT)
''')

while True:
    start_time = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime())
    myrecording = sd.rec(int(fs * duration), samplerate=fs, channels=1)
    sd.wait()
    speech_chunk = myrecording.flatten()
    # 噪声处理
    filtered_chunk = np.convolve(speech_chunk, np.ones(window_size) / window_size, mode='same')
    speech_chunk = filtered_chunk
    is_final = False
    res = model.generate(input=speech_chunk, cache=cache, is_final=is_final, chunk_size=chunk_size,
                         encoder_chunk_look_back=encoder_chunk_look_back,
                         decoder_chunk_look_back=decoder_chunk_look_back)
    text_result=''.join(lazy_pinyin(str(res[0]['text']))).replace(" ", "")
    # 唤醒词
    s1=''.join(lazy_pinyin(str("小爱")))
    print(s1)
    print(text_result)
    if s1 in text_result:
        print("我在，我要做什么")
    cursor.execute("INSERT INTO speech_data (text, time_stamp, batch) VALUES (?,?,?)",
                   (text_result, start_time, 'eerr'))
    conn.commit()

3.效果

4.问题

1.必须有麦克风才能跑起来
2.关于模型包，可以直接从模型社区下载(实在找不到可私信我)
3.最后的效果与你电脑的显卡有直接联系