手搓智能音箱——语音识别及调用大模型回应

一、代码概述

此 Python 代码实现了一个语音交互系统，主要功能为监听唤醒词，在唤醒后接收用户语音问题，利用百度语音识别将语音转换为文本，再调用 DeepSeek API 获取智能回复，最后使用文本转语音功能将回复朗读出来。

二、环境依赖

在运行此代码前，需要安装以下 Python 库：

baidu-aip：用于调用百度语音识别服务。
pyttsx3：实现文本转语音功能。
dotenv：用于加载环境变量。
pyaudio：用于录制音频。
wave：用于处理音频文件。
requests：用于发送 HTTP 请求。
json：用于处理 JSON 数据。

可以使用以下命令进行安装：

bash

pip install baidu-aip pyttsx3 python-dotenv pyaudio requests

三、环境变量配置

需要在项目根目录下创建一个 .env 文件，并添加以下环境变量：

plaintext

BAIDU_APP_ID=your_baidu_app_id
BAIDU_API_KEY=your_baidu_api_key
BAIDU_SECRET_KEY=your_baidu_secret_key
DEEPSEEK_API_KEY=your_deepseek_api_key

请将 your_baidu_app_id、your_baidu_api_key、your_baidu_secret_key 替换为你在百度语音平台申请的应用 ID、API 密钥和 Secret 密钥，将 your_deepseek_api_key 替换为你在 DeepSeek 平台申请的 API 密钥。

四、代码模块详细解释

1. 导入必要的库

python

from aip import AipSpeech
import pyttsx3
import os
from dotenv import load_dotenv
import wave
import pyaudio
import requests
import json
from datetime import datetime

AipSpeech：用于调用百度语音识别服务。
pyttsx3：实现文本转语音功能。
os：用于操作环境变量。
load_dotenv：从 .env 文件中加载环境变量。
wave：用于处理音频文件。
pyaudio：用于录制音频。
requests：用于发送 HTTP 请求。
json：用于处理 JSON 数据。
datetime：虽然代码中未实际使用，但可用于后续添加时间相关功能。

2. 加载环境变量并初始化客户端

python

# Load environment variables
load_dotenv()
APP_ID = os.getenv("BAIDU_APP_ID")
API_KEY = os.getenv("BAIDU_API_KEY")
SECRET_KEY = os.getenv("BAIDU_SECRET_KEY")# Initialize Baidu Speech client
client = AipSpeech(APP_ID, API_KEY, SECRET_KEY)

load_dotenv()：从 .env 文件中加载环境变量。
AipSpeech：使用百度提供的应用 ID、API 密钥和 Secret 密钥初始化百度语音识别客户端。

3. 初始化文本转语音引擎

python

# Initialize text-to-speech
engine = pyttsx3.init()# Set Chinese voice if available, and choose a gentle female voice
voices = engine.getProperty('voices')
for voice in voices:# 检查音色是否为中文，并且音色名称中包含“female”或“woman”等关键词if 'chinese' in voice.name.lower() and ('female' in voice.name.lower() or 'woman' in voice.name.lower()):engine.setProperty('voice', voice.id)break

pyttsx3.init()：初始化文本转语音引擎。
通过遍历可用语音，选择支持中文且为女性的语音。

4. 调用 DeepSeek API 获取智能回复

python

def call_deepseek_api(prompt):"""Call DeepSeek API for intelligent response"""url = "https://api.deepseek.com/v1/chat/completions"headers = {"Content-Type": "application/json","Authorization": f"Bearer {os.getenv('DEEPSEEK_API_KEY')}"}data = {"model": "deepseek-chat","messages": [{"role": "user","content": prompt}],"temperature": 0.7,"max_tokens": 200}try:response = requests.post(url, headers=headers, json=data)response.raise_for_status()result = response.json()return result['choices'][0]['message']['content'].strip()except Exception as e:print(f"Error calling DeepSeek API: {str(e)}")return "抱歉，我暂时无法回答这个问题"

call_deepseek_api 函数：接收用户的问题 prompt，向 DeepSeek API 发送请求，获取智能回复。
requests.post：发送 HTTP POST 请求。
response.raise_for_status()：检查请求是否成功。
若请求失败，捕获异常并返回错误提示信息。

5. 录制音频

python

def record_audio(filename, record_seconds=5):"""Record audio from microphone"""CHUNK = 1024FORMAT = pyaudio.paInt16CHANNELS = 1RATE = 16000p = pyaudio.PyAudio()stream = p.open(format=FORMAT,channels=CHANNELS,rate=RATE,input=True,frames_per_buffer=CHUNK)print("开始录音...")frames = []for _ 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()wf = wave.open(filename, 'wb')wf.setnchannels(CHANNELS)wf.setsampwidth(p.get_sample_size(FORMAT))wf.setframerate(RATE)wf.writeframes(b''.join(frames))wf.close()

record_audio 函数：用于录制音频，将录制的音频保存为指定文件名的 .wav 文件。
pyaudio.PyAudio()：初始化 PyAudio 对象。
p.open：打开音频输入流。
wave.open：打开音频文件进行写入操作。

6. 监听唤醒词

python

def listen_for_wake_word(wake_word="小绿同学"):"""Listen for the wake word using Baidu Speech Recognition"""temp_file = "temp.wav"record_audio(temp_file, 3)with open(temp_file, 'rb') as f:audio_data = f.read()result = client.asr(audio_data, 'wav', 16000, {'dev_pid': 1537  # Mandarin})if result.get('err_no') == 0:text = result['result'][0]if wake_word in text:return Truereturn False

listen_for_wake_word 函数：录制 3 秒音频，使用百度语音识别将音频转换为文本，检查文本中是否包含唤醒词。
client.asr：调用百度语音识别服务。

7. 获取回复

python

def get_response(prompt):"""Generate response using DeepSeek API"""return call_deepseek_api(prompt)

get_response 函数：调用 call_deepseek_api 函数获取智能回复。

8. 文本转语音

python

def speak(text):"""Convert text to speech"""engine.say(text)engine.runAndWait()

speak 函数：使用 pyttsx3 引擎将文本转换为语音并朗读出来。

9. 主函数

python

def main():print("小绿同学已启动，等待唤醒...")while True:if listen_for_wake_word():print("唤醒成功！请说出你的问题...")speak("您好，需要什么帮助")temp_file = "question.wav"record_audio(temp_file, 5)with open(temp_file, 'rb') as f:audio_data = f.read()result = client.asr(audio_data, 'wav', 16000, {'dev_pid': 1537  # Mandarin})if result.get('err_no') == 0:user_input = result['result'][0]print(f"你说: {user_input}")response = get_response(user_input)print(f"小绿同学: {response}")speak(response)else:print("抱歉，我没有听清楚")speak("抱歉，我没有听清楚")if __name__ == "__main__":main()

main 函数：程序的入口点，不断监听唤醒词，唤醒后录制用户问题，进行语音识别，调用 DeepSeek API 获取回复，并将回复朗读出来。

五、注意事项

确保 .env 文件中的环境变量配置正确。
若网络连接不稳定或 DeepSeek API 服务不可用，可能会导致获取回复失败。
录音设备需要正常工作，否则可能无法录制音频。

完整代码

from aip import AipSpeech
import pyttsx3
import os
from dotenv import load_dotenv
import wave
import pyaudio
import requests
import json
from datetime import datetime# Load environment variables
load_dotenv()
APP_ID = os.getenv("BAIDU_APP_ID")
API_KEY = os.getenv("BAIDU_API_KEY")
SECRET_KEY = os.getenv("BAIDU_SECRET_KEY")# Initialize Baidu Speech client
client = AipSpeech(APP_ID, API_KEY, SECRET_KEY)# Initialize text-to-speech
engine = pyttsx3.init()# Set Chinese voice if available, and choose a gentle female voice
voices = engine.getProperty('voices')
for voice in voices:# 检查音色是否为中文，并且音色名称中包含“female”或“woman”等关键词if 'chinese' in voice.name.lower() and ('female' in voice.name.lower() or 'woman' in voice.name.lower()):engine.setProperty('voice', voice.id)break
def call_deepseek_api(prompt):"""Call DeepSeek API for intelligent response"""url = "https://api.deepseek.com/v1/chat/completions"headers = {"Content-Type": "application/json","Authorization": f"Bearer {os.getenv('DEEPSEEK_API_KEY')}"}data = {"model": "deepseek-chat","messages": [{"role": "user","content": prompt}],"temperature": 0.7,"max_tokens": 200}try:response = requests.post(url, headers=headers, json=data)response.raise_for_status()result = response.json()return result['choices'][0]['message']['content'].strip()except Exception as e:print(f"Error calling DeepSeek API: {str(e)}")return "抱歉，我暂时无法回答这个问题"def record_audio(filename, record_seconds=5):"""Record audio from microphone"""CHUNK = 1024FORMAT = pyaudio.paInt16CHANNELS = 1RATE = 16000p = pyaudio.PyAudio()stream = p.open(format=FORMAT,channels=CHANNELS,rate=RATE,input=True,frames_per_buffer=CHUNK)print("开始录音...")frames = []for _ 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()wf = wave.open(filename, 'wb')wf.setnchannels(CHANNELS)wf.setsampwidth(p.get_sample_size(FORMAT))wf.setframerate(RATE)wf.writeframes(b''.join(frames))wf.close()def listen_for_wake_word(wake_word="小绿"):"""Listen for the wake word using Baidu Speech Recognition"""temp_file = "temp.wav"record_audio(temp_file, 3)with open(temp_file, 'rb') as f:audio_data = f.read()result = client.asr(audio_data, 'wav', 16000, {'dev_pid': 1537  # Mandarin})if result.get('err_no') == 0:text = result['result'][0]if wake_word in text:return Truereturn Falsedef get_response(prompt):"""Generate response using DeepSeek API"""return call_deepseek_api(prompt)def speak(text):"""Convert text to speech"""engine.say(text)engine.runAndWait()def main():print("小绿已启动，等待唤醒...")while True:if listen_for_wake_word():print("唤醒成功！请说出你的问题...")speak("您好，需要什么帮助")temp_file = "question.wav"record_audio(temp_file, 5)with open(temp_file, 'rb') as f:audio_data = f.read()result = client.asr(audio_data, 'wav', 16000, {'dev_pid': 1537  # Mandarin})if result.get('err_no') == 0:user_input = result['result'][0]print(f"你说: {user_input}")response = get_response(user_input)print(f"小绿: {response}")speak(response)else:print("抱歉，我没有听清楚")speak("抱歉，我没有听清楚")if __name__ == "__main__":main()

修改了一些，调用百度进行了语音合成输出，回答更自然一些

from aip import AipSpeech
import os
from dotenv import load_dotenv
import wave
import pyaudio
import requests
import json
from datetime import datetime
import pygame  # 用于播放音频
import tempfile# Load environment variables
load_dotenv()
APP_ID = os.getenv("BAIDU_APP_ID")
API_KEY = os.getenv("BAIDU_API_KEY")
SECRET_KEY = os.getenv("BAIDU_SECRET_KEY")# Initialize Baidu Speech client
client = AipSpeech(APP_ID, API_KEY, SECRET_KEY)def call_deepseek_api(prompt):"""Call DeepSeek API for intelligent response"""url = "https://api.deepseek.com/v1/chat/completions"headers = {"Content-Type": "application/json","Authorization": f"Bearer {os.getenv('DEEPSEEK_API_KEY')}"}data = {"model": "deepseek-chat","messages": [{"role": "user","content": prompt}],"temperature": 0.7,"max_tokens": 200}try:response = requests.post(url, headers=headers, json=data)response.raise_for_status()result = response.json()return result['choices'][0]['message']['content'].strip()except Exception as e:print(f"Error calling DeepSeek API: {str(e)}")return "抱歉，我暂时无法回答这个问题"def record_audio(filename, record_seconds=5):"""Record audio from microphone"""CHUNK = 1024FORMAT = pyaudio.paInt16CHANNELS = 1RATE = 16000p = pyaudio.PyAudio()stream = p.open(format=FORMAT,channels=CHANNELS,rate=RATE,input=True,frames_per_buffer=CHUNK)print("开始录音...")frames = []for _ 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()wf = wave.open(filename, 'wb')wf.setnchannels(CHANNELS)wf.setsampwidth(p.get_sample_size(FORMAT))wf.setframerate(RATE)wf.writeframes(b''.join(frames))wf.close()def listen_for_wake_word(wake_word="小绿"):"""Listen for the wake word using Baidu Speech Recognition"""temp_file = "temp.wav"record_audio(temp_file, 3)with open(temp_file, 'rb') as f:audio_data = f.read()result = client.asr(audio_data, 'wav', 16000, {'dev_pid': 1537  # Mandarin})if result.get('err_no') == 0:text = result['result'][0]if wake_word in text:return Truereturn Falsedef get_response(prompt):"""Generate response using DeepSeek API"""return call_deepseek_api(prompt)def speak(text):"""Convert text to speech using Baidu TTS"""result = client.synthesis(text, 'zh', 1, {'vol': 5,  # 音量，取值0-15，默认为5中音量'spd': 4,  # 语速，取值0-9，默认为5中语速'pit': 5,  # 音调，取值0-9，默认为5中语调'per': 4   # 发音人选择，0为女声，1为男声，3为度逍遥，4为度丫丫})if not isinstance(result, dict):# 使用临时文件with tempfile.NamedTemporaryFile(suffix=".mp3", delete=False) as f:f.write(result)temp_path = f.name# 播放语音文件pygame.mixer.init()pygame.mixer.music.load(temp_path)pygame.mixer.music.play()# 等待播放完成while pygame.mixer.music.get_busy():pygame.time.Clock().tick(10)  # 控制循环频率，避免占用过多 CPU# 强制释放资源pygame.mixer.music.stop()pygame.mixer.quit()# 播放完成后删除临时文件os.remove(temp_path)else:print(f"语音合成失败: {result}")def main():print("小绿已启动，等待唤醒...")while True:if listen_for_wake_word():print("您好，需要什么帮助....")speak("您好，需要什么帮助")temp_file = "question.wav"record_audio(temp_file, 5)with open(temp_file, 'rb') as f:audio_data = f.read()result = client.asr(audio_data, 'wav', 16000, {'dev_pid': 1537  # Mandarin})if result.get('err_no') == 0:user_input = result['result'][0]print(f"你说: {user_input}")response = get_response(user_input)print(f"小绿: {response}")speak(response)else:print("抱歉，我没有听清楚")speak("抱歉，我没有听清楚")if __name__ == "__main__":main()