python调用麦克风和扬声器，并调用百度实时语音转文字_python麦克风

# [1]导入必要的模块和配置百度的 SDK
import time
import queue
import sounddevice as sd
import numpy as np
from aip import AipSpeech
import sys
 
# 百度云配置信息
APP_ID = ''  # 替换为实际的 APP ID
API_KEY = ''  # 替换为实际的 API KEY
SECRET_KEY = ''  # 替换为实际的 SECRET KEY
 
client = AipSpeech(APP_ID, API_KEY, SECRET_KEY)
 
# Queue to hold the recorded audio data
audio_queue = queue.Queue()
speaker_queue = queue.Queue()
 
# Callback function to capture audio data from microphone
def audio_callback(indata, frames, time, status):
    if status:
        print(status, file=sys.stderr)
    audio_queue.put(indata.copy())
 
# Callback function to capture audio data from speaker
def speaker_callback(indata, frames, time, status):
    if status:
        print(status, file=sys.stderr)
    speaker_queue.put(indata.copy())
# [2]实现实时语音识别类
class RealTimeSpeechRecognizer:
    def __init__(self, client, name):
        self.client = client
        self.name = name
 
    def send_audio(self, audio_data):
        result = self.client.asr(audio_data, 'pcm', 16000, {
            'dev_pid': 1537,
        })
        if result.get('err_no') == 0:
            print(f"{self.name} 识别结果: {result['result']}")
        else:
            print(f"{self.name} 错误: {result['err_msg']}")
 
# 调用百度的语音转文字的接口
def recognize_speech(audio_data, recognizer):
    audio_data = np.concatenate(audio_data)
    recognizer.send_audio(audio_data.tobytes())
# [3]开始音频流并处理音频数据
def start_audio_stream(mic_recognizer, speaker_recognizer, speaker_device_index):
    with sd.InputStream(callback=audio_callback, channels=1, samplerate=16000, dtype='int16') as mic_stream, \
            sd.InputStream(callback=speaker_callback, channels=1, samplerate=16000, dtype='int16', device=speaker_device_index) as spk_stream:
        print("Recording audio... Press Ctrl+C to stop.")
        mic_audio_buffer = []
        speaker_audio_buffer = []
        try:
            while True:
                while not audio_queue.empty():
                    mic_audio_buffer.append(audio_queue.get())
                while not speaker_queue.empty():
                    speaker_audio_buffer.append(speaker_queue.get())
 
                if len(mic_audio_buffer) >= 10:
                    recognize_speech(mic_audio_buffer, mic_recognizer)
                    mic_audio_buffer = []  # Clear buffer after sending
 
                if len(speaker_audio_buffer) >= 10:
                    recognize_speech(speaker_audio_buffer, speaker_recognizer)
                    speaker_audio_buffer = []  # Clear buffer after sending
 
                time.sleep(0.1)
        except KeyboardInterrupt:
            print("Stopping audio recording.")
# [4]主程序入口
if __name__ == "__main__":
    speaker_device_index = 8  # 使用 pulse 设备（索引 8）来捕获扬声器输出
 
    mic_recognizer = RealTimeSpeechRecognizer(client, "麦克风接收：")
    speaker_recognizer = RealTimeSpeechRecognizer(client, "扬声器接收：")
 
    start_audio_stream(mic_recognizer, speaker_recognizer, speaker_device_index)

一、实时的短语音识别场景

在某些应用场景中，可能需要同时捕获麦克风和扬声器的音频数据，例如以下几种情况：

1. 实时翻译和转录会议

在一个会议或对话中，你可能需要捕获：

• 麦克风音频：捕获发言者的声音，进行实时转录。
• 扬声器音频：捕获对话的另一方通过扬声器播放的声音，进行实时转录。

这样可以同时转录双方的发言，提供完整的对话记录。

2. 语言学习和教学

在语言学习的应用中，教师可能会播放音频材料，而学生则通过麦克风回答。捕获这两种音频数据可以帮助：

• 麦克风音频：捕获学生的回答和发言。
• 扬声器音频：捕获教师播放的音频材料。

通过同时转录这两种音频，可以对学生的发言和教师播放的材料进行分析和评估。

3. 语音控制系统

在语音控制系统中，系统可能需要捕获：

• 麦克风音频：捕获用户的语音指令。
• 扬声器音频：捕获系统播放的反馈音，确认系统是否正确播放了反馈信息。

这有助于确保系统对用户的指令做出了正确的反馈。

4. 电话会议录音

在电话会议中，可能需要捕获：

• 麦克风音频：捕获本地发言者的声音。
• 扬声器音频：捕获远程参与者通过扬声器播放的声音。

这样可以完整地记录整个会议过程。

基于上述场景，我的代码实现了同时捕获麦克风和扬声器的音频数据，并分别进行转录。

二、实现步骤

1. 导入必要的模块和配置百度的 SDK

import time
import queue
import sounddevice as sd
import numpy as np
from aip import AipSpeech
import sys
 
# 百度云配置信息
APP_ID = '你的 App ID'  # 替换为实际的 APP ID
API_KEY = '你的 Api Key'  # 替换为实际的 API KEY
SECRET_KEY = '你的 Secret Key'  # 替换为实际的 SECRET KEY
 
client = AipSpeech(APP_ID, API_KEY, SECRET_KEY)
 
# Queue to hold the recorded audio data
audio_queue = queue.Queue()
speaker_queue = queue.Queue()
 
# Callback function to capture audio data from microphone
def audio_callback(indata, frames, time, status):
    if status:
        print(status, file=sys.stderr)
    audio_queue.put(indata.copy())
 
# Callback function to capture audio data from speaker
def speaker_callback(indata, frames, time, status):
    if status:
        print(status, file=sys.stderr)
    speaker_queue.put(indata.copy())

2. 实现实时语音识别类

class RealTimeSpeechRecognizer:
    def __init__(self, client, name):
        self.client = client
        self.name = name
 
    def send_audio(self, audio_data):
        result = self.client.asr(audio_data, 'pcm', 16000, {
            'dev_pid': 1537,
        })
        if result.get('err_no') == 0:
            print(f"{self.name} 识别结果: {result['result']}")
        else:
            print(f"{self.name} 错误: {result['err_msg']}")
 
# 调用百度的语音转文字的接口
def recognize_speech(audio_data, recognizer):
    audio_data = np.concatenate(audio_data)
    recognizer.send_audio(audio_data.tobytes())

3. 开始音频流并处理音频数据

def start_audio_stream(mic_recognizer, speaker_recognizer, speaker_device_index):
    with sd.InputStream(callback=audio_callback, channels=1, samplerate=16000, dtype='int16') as mic_stream, \
            sd.InputStream(callback=speaker_callback, channels=1, samplerate=16000, dtype='int16', device=speaker_device_index) as spk_stream:
        print("Recording audio... Press Ctrl+C to stop.")
        mic_audio_buffer = []
        speaker_audio_buffer = []
        try:
            while True:
                while not audio_queue.empty():
                    mic_audio_buffer.append(audio_queue.get())
                while not speaker_queue.empty():
                    speaker_audio_buffer.append(speaker_queue.get())
 
                if len(mic_audio_buffer) >= 10:
                    recognize_speech(mic_audio_buffer, mic_recognizer)
                    mic_audio_buffer = []  # Clear buffer after sending
 
                if len(speaker_audio_buffer) >= 10:
                    recognize_speech(speaker_audio_buffer, speaker_recognizer)
                    speaker_audio_buffer = []  # Clear buffer after sending
 
                time.sleep(0.1)
        except KeyboardInterrupt:
            print("Stopping audio recording.")

4. 主程序入口

if __name__ == "__main__":
    speaker_device_index = 8  # 使用 pulse 设备（索引 8）来捕获扬声器输出
 
    mic_recognizer = RealTimeSpeechRecognizer(client, "麦克风接收：")
    speaker_recognizer = RealTimeSpeechRecognizer(client, "扬声器接收：")
 
    start_audio_stream(mic_recognizer, speaker_recognizer, speaker_device_index)