guoyu/baidu_speech_server.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-
"""
百度语音识别服务（超级简单，不需要ffmpeg）
免费额度：每天50000次
"""

from flask import Flask, request, jsonify
from flask_cors import CORS
import base64
import json
from difflib import SequenceMatcher

app = Flask(__name__)
CORS(app)

# 百度语音识别配置（需要申请）
# 免费申请地址：https://console.bce.baidu.com/ai/#/ai/speech/overview/index
BAIDU_APP_ID = "你的APP_ID"  # ← 需要替换
BAIDU_API_KEY = "你的API_KEY"  # ← 需要替换
BAIDU_SECRET_KEY = "你的SECRET_KEY"  # ← 需要替换

# 懒加载客户端
asr_client = None

def get_asr_client():
    """获取百度语音识别客户端"""
    global asr_client
    if asr_client is None:
        try:
            from aip import AipSpeech
            asr_client = AipSpeech(BAIDU_APP_ID, BAIDU_API_KEY, BAIDU_SECRET_KEY)
            print("✓ 百度语音客户端初始化成功")
        except ImportError:
            print("✗ 未安装百度SDK，请运行: pip install baidu-aip")
            return None
        except Exception as e:
            print(f"✗ 初始化失败: {str(e)}")
            return None
    return asr_client

def recognize_audio_baidu(audio_data, format='mp3'):
    """使用百度API识别音频"""
    try:
        client = get_asr_client()
        if not client:
            return None, "百度语音客户端未初始化"
        
        # 百度API识别
        result = client.asr(audio_data, format, 16000, {
            'dev_pid': 1537,  # 中文普通话
        })
        
        if result['err_no'] == 0:
            text = ''.join(result['result'])
            return text, None
        else:
            return None, f"识别失败: {result.get('err_msg', '未知错误')}"
            
    except Exception as e:
        return None, str(e)

def calculate_similarity(text1, text2):
    """计算文本相似度（0-100分）"""
    if not text1 or not text2:
        return 0
    
    text1 = text1.replace(" ", "")
    text2 = text2.replace(" ", "")
    
    if not text1 or not text2:
        return 0
    
    similarity = SequenceMatcher(None, text1, text2).ratio()
    return round(similarity * 100, 2)

@app.route('/api/speech/recognize', methods=['POST'])
def recognize():
    """语音识别接口"""
    try:
        # 检查文件
        if 'audio' not in request.files:
            return jsonify({
                'code': 400,
                'msg': '未上传音频文件'
            }), 400
        
        audio_file = request.files['audio']
        reference_text = request.form.get('referenceText', '')
        
        # 读取音频数据
        audio_data = audio_file.read()
        
        print(f"收到音频: {len(audio_data)} bytes")
        print(f"参考文本: {reference_text}")
        
        # 识别音频（百度API自动处理格式）
        recognized_text, error = recognize_audio_baidu(audio_data, format='mp3')
        
        if error:
            return jsonify({
                'code': 500,
                'msg': f'识别失败: {error}'
            }), 500
        
        if not recognized_text:
            return jsonify({
                'code': 500,
                'msg': '未识别到有效语音'
            }), 500
        
        # 计算评分
        score = calculate_similarity(recognized_text, reference_text)
        pronunciation_score = max(0, score - 5)
        fluency_score = max(0, score - 3)
        
        print(f"识别结果: {recognized_text}")
        print(f"相似度: {score}分")
        
        return jsonify({
            'code': 200,
            'msg': '成功',
            'data': {
                'recognizedText': recognized_text,
                'score': score,
                'pronunciationScore': pronunciation_score,
                'fluencyScore': fluency_score,
                'status': 'completed'
            }
        })
        
    except Exception as e:
        print(f"处理错误: {str(e)}")
        return jsonify({
            'code': 500,
            'msg': f'处理失败: {str(e)}'
        }), 500

@app.route('/api/speech/health', methods=['GET'])
def health():
    """健康检查"""
    client = get_asr_client()
    return jsonify({
        'code': 200,
        'msg': '服务正常',
        'data': {
            'engine': 'baidu',
            'client_ready': client is not None
        }
    })

if __name__ == '__main__':
    print("=" * 50)
    print("百度语音识别服务（超简单，无需ffmpeg）")
    print("=" * 50)
    print("")
    print("1. 安装依赖: pip install baidu-aip")
    print("2. 申请百度API: https://console.bce.baidu.com/ai/#/ai/speech/overview/index")
    print("3. 填写 APP_ID, API_KEY, SECRET_KEY")
    print("")
    
    if BAIDU_APP_ID == "你的APP_ID":
        print("⚠️  请先配置百度API密钥！")
        print("")
    
    print("=" * 50)
    print("服务启动成功！")
    print("访问地址: http://localhost:5000")
    print("=" * 50)
    print("")
    
    app.run(host='0.0.0.0', port=5000, debug=False)