# 1、核心庫(kù)安裝
pip install SpeechRecognition

# 2、該庫(kù)處理麥克風(fēng)實(shí)時(shí)語(yǔ)音需要依賴`PyAudio`：
pip install pyaudio
# 若安裝失?。╓indows 常見）：先下載對(duì)應(yīng) Python 版本的PyAudio whl 文件（Unofficial Windows Binaries），再本地安裝
pip install PyAudio-0.2.13-cp310-cp310-win_amd64.whl

# 3、Mac 用戶：先裝底層依賴再裝 PyAudio
brew install portaudio
pip install pyaudio

# 4、若需離線識(shí)別，額外安裝
pip install pocketsphinx

import speech_recognition as sr

def microphone_recognition():
    # 1. 創(chuàng)建Recognizer實(shí)例
    r = sr.Recognizer()
    
    # 2. 獲取麥克風(fēng)設(shè)備（默認(rèn)麥克風(fēng)）
    with sr.Microphone() as source:
        print("正在校準(zhǔn)環(huán)境噪聲，請(qǐng)保持安靜...")
        # 校準(zhǔn)噪聲：監(jiān)聽1秒，獲取背景噪聲水平（關(guān)鍵步驟，否則識(shí)別準(zhǔn)確率低）
        r.adjust_for_ambient_noise(source, duration=1)
        print("校準(zhǔn)完成，請(qǐng)開始說(shuō)話（說(shuō)完后會(huì)自動(dòng)識(shí)別）...")
        
        # 3. 監(jiān)聽麥克風(fēng)音頻（timeout=5表示5秒內(nèi)無(wú)語(yǔ)音則超時(shí)）
        audio = r.listen(source, timeout=5, phrase_time_limit=10)  # phrase_time_limit限制單次語(yǔ)音最長(zhǎng)10秒
    
    # 4. 調(diào)用Google API識(shí)別語(yǔ)音
    try:
        print("正在識(shí)別...")
        # language="zh-CN"指定中文識(shí)別，默認(rèn)是英文
        # 若識(shí)別英文可改為`language="en-US"`。
        text = r.recognize_google(audio, language="zh-CN")
        print(f"你說(shuō)的內(nèi)容：{text}")
    except sr.WaitTimeoutError:
        print("錯(cuò)誤：超過(guò)5秒未檢測(cè)到語(yǔ)音輸入")
    except sr.UnknownValueError:
        print("錯(cuò)誤：無(wú)法識(shí)別你說(shuō)的內(nèi)容（語(yǔ)音模糊/無(wú)有效語(yǔ)音）")
    except sr.RequestError as e:
        print(f"錯(cuò)誤：調(diào)用Google API失敗 → {e}")

if __name__ == "__main__":
    microphone_recognition()

import speech_recognition as sr

def audio_file_recognition(file_path):
    # 1. 創(chuàng)建Recognizer實(shí)例
    r = sr.Recognizer()
    
    # 2. 讀取音頻文件
    with sr.AudioFile(file_path) as source:
        # 讀取完整音頻（也可指定時(shí)長(zhǎng)：r.record(source, duration=5) 讀取前5秒）
        audio = r.record(source)
    
    # 3. 識(shí)別音頻
    try:
        text = r.recognize_google(audio, language="zh-CN")
        print(f"音頻內(nèi)容：{text}")
    except sr.UnknownValueError:
        print("錯(cuò)誤：無(wú)法識(shí)別音頻內(nèi)容")
    except sr.RequestError as e:
        print(f"錯(cuò)誤：API調(diào)用失敗 → {e}")

if __name__ == "__main__":
    # 替換為你的音頻文件路徑（建議用wav格式）
    audio_file_recognition("test_audio.wav")

pip install pydub
# Windows需下載ffmpeg并配置環(huán)境變量，Mac：brew install ffmpeg

import speech_recognition as sr
from pydub import AudioSegment

# 轉(zhuǎn)換MP3為WAV
audio = AudioSegment.from_mp3("test_audio.mp3")
audio.export("test_audio_wav.wav", format="wav")

# 調(diào)用上面的識(shí)別函數(shù)
audio_file_recognition("test_audio_wav.wav")

import speech_recognition as sr

def offline_recognition():
    r = sr.Recognizer()
    with sr.Microphone() as source:
        r.adjust_for_ambient_noise(source, duration=1)
        print("離線模式，請(qǐng)說(shuō)話...")
        audio = r.listen(source)
    
    try:
        # 離線識(shí)別（中文需額外配置語(yǔ)言包，默認(rèn)英文）
        text = r.recognize_sphinx(audio, language="zh-CN")
        print(f"離線識(shí)別結(jié)果：{text}")
    except sr.UnknownValueError:
        print("錯(cuò)誤：離線識(shí)別失敗（語(yǔ)音無(wú)法解析）")
    except sr.RequestError as e:
        print(f"錯(cuò)誤：Sphinx引擎初始化失敗 → {e}")

if __name__ == "__main__":
    offline_recognition()

   pip install baidu-aip

import speech_recognition as sr
from aip import AipSpeech

# 替換為你的百度應(yīng)用信息
APP_ID = "你的APP_ID"
API_KEY = "你的API_KEY"
SECRET_KEY = "你的SECRET_KEY"

def baidu_recognition(audio_data):
    # 初始化百度語(yǔ)音客戶端
    client = AipSpeech(APP_ID, API_KEY, SECRET_KEY)
    
    # 將AudioData轉(zhuǎn)為百度要求的格式（pcm/16k/16位/單聲道）
    pcm_data = audio_data.get_raw_data(convert_rate=16000, convert_width=2)
    
    # 調(diào)用百度API識(shí)別
    result = client.asr(pcm_data, "pcm", 16000, {
        "dev_pid": 1536,  # 1536=普通話（支持簡(jiǎn)單的英文），1537=英語(yǔ)
    })
    
    if result["err_no"] == 0:
        return "".join(result["result"])
    else:
        raise Exception(f"百度API錯(cuò)誤：{result['err_msg']}")

# 主流程：麥克風(fēng)采集+百度識(shí)別
if __name__ == "__main__":
    r = sr.Recognizer()
    with sr.Microphone(sample_rate=16000) as source:  # 百度要求16k采樣率
        r.adjust_for_ambient_noise(source, duration=1)
        print("請(qǐng)說(shuō)話（百度語(yǔ)音識(shí)別）...")
        audio = r.listen(source)
    
    try:
        text = baidu_recognition(audio)
        print(f"百度識(shí)別結(jié)果：{text}")
    except Exception as e:
        print(f"識(shí)別失?。簕e}")