Python声音识别：从技术原理到实战落地

# 安装核心依赖库
pip install SpeechRecognition pyaudio librosa

import speech_recognition as sr

# 初始化识别器
recognizer = sr.Recognizer()

# 使用麦克风作为音频源
with sr.Microphone() as source:
    print("请开始说话（5秒内）...")
    # 调整环境噪音（关键步骤，提升识别精度）
    recognizer.adjust_for_ambient_noise(source, duration=0.5)
    # 捕获音频数据
    audio = recognizer.listen(source, timeout=5)
    
    try:
        print("正在识别...")
        # 调用Google Web Speech API识别中文
        text = recognizer.recognize_google(audio, language='zh-CN')
        print(f"识别结果：{text}")
    except sr.UnknownValueError:
        print("无法识别音频内容")
    except sr.RequestError as e:
        print(f"识别服务请求失败：{e}")

# 识别本地音频文件
with sr.AudioFile("audio.wav") as source:
    audio = recognizer.record(source)  # 读取完整音频
    # 识别音频
    text = recognizer.recognize_google(audio, language='zh-CN')
    print(f"文件识别结果：{text}")

import speech_recognition as sr
from pydub import AudioSegment
from aip import AipSpeech

# 1. MP3转WAV（需安装ffmpeg）
def mp3_to_wav(mp3_path, wav_path):
    audio = AudioSegment.from_mp3(mp3_path)
    audio.export(wav_path, format="wav")

# 2. 百度AI语音识别配置（需在百度云申请账号获取）
APP_ID = "你的APP_ID"
API_KEY = "你的API_KEY"
SECRET_KEY = "你的SECRET_KEY"
client = AipSpeech(APP_ID, API_KEY, SECRET_KEY)

# 3. 读取音频文件并转Base64（百度API要求）
def get_audio_content(wav_path):
    with open(wav_path, "rb") as f:
        return f.read()

# 4. 长音频分段识别（每30秒一段）
def batch_recognize_meeting(wav_path):
    audio = AudioSegment.from_wav(wav_path)
    chunk_duration = 30 * 1000  # 30秒（毫秒）
    chunks = [audio[i:i+chunk_duration] for i in range(0, len(audio), chunk_duration)]
    
    full_text = ""
    for i, chunk in enumerate(chunks):
        chunk_path = f"chunk_{i}.wav"
        chunk.export(chunk_path, format="wav")
        # 调用百度AI识别长音频
        result = client.asr(get_audio_content(chunk_path), "wav", 16000, {"dev_pid": 1537})
        if result["err_no"] == 0:
            full_text += "".join(result["result"]) + "\n"
    # 保存为会议纪要
    with open("会议纪要.txt", "w", encoding="utf-8") as f:
        f.write(full_text)
    print("会议纪要生成完成！")

# 执行流程
mp3_to_wav("meeting.mp3", "meeting.wav")
batch_recognize_meeting("meeting.wav")

import speech_recognition as sr
import os
import datetime

def execute_command(command):
    """根据指令执行对应操作"""
    if "打开浏览器" in command:
        os.startfile("C:\\Program Files\\Google\\Chrome\\Application\\chrome.exe")  # 适配本地路径
        print("已打开浏览器")
    elif "查询时间" in command:
        current_time = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
        print(f"当前时间：{current_time}")
    elif "退出" in command:
        print("语音助手已退出")
        return False
    return True

def voice_assistant():
    recognizer = sr.Recognizer()
    with sr.Microphone() as source:
        print("智能语音助手已启动，等待指令...")
        recognizer.adjust_for_ambient_noise(source, duration=0.3)
        while True:
            audio = recognizer.listen(source, timeout=10, phrase_time_limit=5)
            try:
                command = recognizer.recognize_google(audio, language='zh-CN')
                print(f"你说：{command}")
                if not execute_command(command):
                    break
            except sr.UnknownValueError:
                print("未听清，请重复指令")
            except sr.RequestError:
                print("https://m.jixing.net/shengxiao/")

# 启动助手
voice_assistant()

import librosa
import soundfile as sf

# 过滤低频噪音
y, sr = librosa.load("noisy_audio.wav", sr=16000)
y_filtered = librosa.effects.trim(y, top_db=20)[0]  # 去除静音和低频噪音
sf.write("clean_audio.wav", y_filtered, sr)

import librosa
import matplotlib.pyplot as plt

# 提取MFCC特征（声音识别核心特征）
y, sr = librosa.load("audio.wav", sr=16000)
mfccs = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=13)  # 13维MFCC特征
# 可视化MFCC特征
plt.figure(figsize=(10, 4))
librosa.display.specshow(mfccs, x_axis="time", sr=sr)
plt.colorbar()
plt.title("MFCC特征图")
plt.tight_layout()
plt.show()