春色将阑，莺声渐老，红英落尽青梅小。画堂人静雨蒙蒙，屏山半掩余香袅。
密约沉沉，离情杳杳，菱花尘满慵将照。倚楼无语欲销魂，长空黯淡连芳草。

首先安装python包

pip3 install baidu-aip

再注册百度AI得到相关参数

https://ai.baidu.com/

一、语音合成

from aip import AipSpeech APP_ID = '14446020' API_KEY = 'GnaoLWrIiTKP10disiDHMiNZ' SECRET_KEY = 'FYaMNBsH5NFsgWcRsyBfaHDV70MvvE6u' #实例化AipSpeech对象 client = AipSpeech(APP_ID, API_KEY, SECRET_KEY) #调用方法语音合成 result = client.synthesis("欢迎来到王者荣耀", options={ "per": 1, "spd": 9, "pit": 9, "vol": 15, }) if not isinstance(result, dict): with open('auido.mp3', 'wb') as f: f.write(result) 

二、语音识别

from aip import AipSpeech import os APP_ID = '14446020' API_KEY = 'GnaoLWrIiTKP10disiDHMiNZ' SECRET_KEY = 'FYaMNBsH5NFsgWcRsyBfaHDV70MvvE6u' client = AipSpeech(APP_ID, API_KEY, SECRET_KEY) def get_file_content(filePath): cmd_str = f"ffmpeg -y -i {filePath} -acodec pcm_s16le -f s16le -ac 1 -ar 16000 {filePath}.pcm" os.system(cmd_str) with open(f"{filePath}.pcm", 'rb') as fp: return fp.read() res = client.asr(speech=get_file_content("auido.mp3"),options={ "dev_pid":1536, }) print(res["result"][0]) 

这里用到一个格式转换的软件，百度地址：
https://pan.baidu.com/s/1MadxSh-A0Pzo1Su_wKdktQ 提取码：x5xi
固定的格式转换命令：（需要将bin文件添加环境变量，在cmd中执行）
ffmpeg -y -i filePath -acodec pcm_s16le -f s16le -ac 1 -ar 16000 filePath.pcm

三、短文本相似度

from aip import AipNlp APP_ID = '14446020' API_KEY = 'GnaoLWrIiTKP10disiDHMiNZ' SECRET_KEY = 'FYaMNBsH5NFsgWcRsyBfaHDV70MvvE6u' client = AipNlp(APP_ID,API_KEY,SECRET_KEY) ret = client.simnet("你今年几岁了？","多大年龄了？") print(ret) 

{'log_id': 4545309161914786697, 'texts': {'text_2': '多大年龄了？', 'text_1': '你今年几岁了？'}, 'score': 0.742316} score 是两个测试的短文本相似度，一般大于0.72的两个短文本的意思是相似的句子！

四、代码实现对接图灵

import requests def tuling_test(question): url = "http://openapi.tuling123.com/openapi/api/v2" data = { "reqType":0, "perception": { "inputText": { "text": question }, "inputImage": { }, }, "userInfo": { "apiKey": "2f4e809b8b3049ce82a6b4787bad65bb", "userId": "wangjifei" } } return requests.post(url=url,json=data).json() ret = tuling_test("心情不好") print(ret.get("results")[0]["values"]["text"]) 

五、简单实现人机交流

• 基本步骤：
  用户录制音频---传入函数---格式转化---语音识别---匹配答案---语音合成---语音文件流写入文件---os执行文件---删除文件

from aip import AipSpeech from aip import AipNlp from uuid import uuid4 import os import requests import time APP_ID = '14446007' API_KEY = 'QrQWLLg5a8qld7Qty7avqCGC' SECRET_KEY = 'O5mE31LSl17hm8NRYyf9PwlE5Byqm0nr' client = AipSpeech(APP_ID, API_KEY, SECRET_KEY) nlp_client = AipNlp(APP_ID, API_KEY, SECRET_KEY) def tuling_test(question): """接入图灵，为问题匹配答案""" url = "http://openapi.tuling123.com/openapi/api/v2" data = { "reqType": 0, "perception": { "inputText": { "text": question }, "inputImage": { }, }, "userInfo": { "apiKey": "2f4e809b8b3049ce82a6b4787bad65bb", "userId": "wangjifei" } } ret = requests.post(url=url, json=data).json() return ret.get("results")[0]["values"]["text"] def get_file_content(filePath): """音频的格式转换""" cmd_str = f"ffmpeg -y -i {filePath} -acodec pcm_s16le -f s16le -ac 1 -ar 16000 {filePath}.pcm" os.system(cmd_str) with open(f"{filePath}.pcm", 'rb') as fp: return fp.read() def custom_reply(text): """根据问题得到相应的答案，可以通过短文本相似来自定义，也可以调用图灵问题库""" if nlp_client.simnet("你叫什么名字", text).get("score") >= 0.72: return "我不能告诉你" return tuling_test(text) def learn_say(file_name): """机器人学说话""" # 语音识别成文字 res = client.asr(speech=get_file_content(file_name), options={ "dev_pid": 1536, }) os.remove(f"{file_name}.pcm") text = res.get("result")[0] # 根据问题得到相关答案 text1 = custom_reply(text) # 答案语音合成 res_audio = client.synthesis(text1, options={ "vol": 8, "pit": 8, "spd": 5, "per": 4 }) # 通过uuid 生成文件名 ret_file_name = f"{uuid4()}.mp3" # 将生成的语音流写入文件中 with open(ret_file_name, "wb") as f: f.write(res_audio) # 执行音频文件 ret = os.system(ret_file_name) time.sleep(2) os.remove(ret_file_name) if __name__ == '__main__': learn_say("auido.m4a") 

六、网页版智能机器人对话

flask_ws.py

from flask import Flask, request, render_template from uuid import uuid4 from geventwebsocket.websocket import WebSocket from gevent.pywsgi import WSGIServer from geventwebsocket.handler import WebSocketHandler from learn_say import learn_say app = Flask(__name__) # type:Flask @app.route("/ws") def ws(): user_socket = request.environ.get("wsgi.websocket") # type:WebSocket while True: msg = user_socket.receive() q_file_name = f"{uuid4()}.wav" with open(q_file_name, "wb") as f: f.write(msg) ret_file_name = learn_say(q_file_name) user_socket.send(ret_file_name) if __name__ == '__main__': http_serv = WSGIServer(("127.0.0.1", 8006), app, handler_class=WebSocketHandler) http_serv.serve_forever() 

flask_app.py

from flask import Flask, request, render_template, send_file app = Flask(__name__) # type:Flask @app.route("/index") def index(): return render_template("index.html") @app.route("/get_audio/<audio_name>") def get_audio(audio_name): return send_file(audio_name) if __name__ == '__main__': app.run("127.0.0.1", 8008, debug=True) 

learn_say.py

from aip import AipSpeech from aip import AipNlp from uuid import uuid4 import os import requests import time APP_ID = '14446007' API_KEY = 'QrQWLLg5a8qld7Qty7avqCGC' SECRET_KEY = 'O5mE31LSl17hm8NRYyf9PwlE5Byqm0nr' client = AipSpeech(APP_ID, API_KEY, SECRET_KEY) nlp_client = AipNlp(APP_ID, API_KEY, SECRET_KEY) def tuling_test(question): """接入图灵，为问题匹配答案""" url = "http://openapi.tuling123.com/openapi/api/v2" data = { "reqType": 0, "perception": { "inputText": { "text": question }, "inputImage": { }, }, "userInfo": { "apiKey": "2f4e809b8b3049ce82a6b4787bad65bb", "userId": "wangjifei" } } ret = requests.post(url=url, json=data).json() return ret.get("results")[0]["values"]["text"] def get_file_content(filePath): """音频的格式转换""" cmd_str = f"ffmpeg -y -i {filePath} -acodec pcm_s16le -f s16le -ac 1 -ar 16000 {filePath}.pcm" os.system(cmd_str) with open(f"{filePath}.pcm", 'rb') as fp: return fp.read() def custom_reply(text): """根据问题得到相应的答案，可以通过短文本相似来自定义，也可以调用图灵问题库""" if nlp_client.simnet("你叫什么名字", text).get("score") >= 0.72: return "我不能告诉你" return tuling_test(text) def learn_say(file_name): """机器人学说话""" # 语音识别成文字 res = client.asr(speech=get_file_content(file_name), options={ "dev_pid": 1536, }) os.remove(file_name) os.remove(f"{file_name}.pcm") text = res.get("result")[0] # 根据问题得到相关答案 text1 = custom_reply(text) # 答案语音合成 res_audio = client.synthesis(text1, options={ "vol": 8, "pit": 8, "spd": 5, "per": 4 }) # 通过uuid 生成文件名 ret_file_name = f"{uuid4()}.mp3" # 将生成的语音流写入文件中 with open(ret_file_name, "wb") as f: f.write(res_audio) return ret_file_name 

index.html

<!DOCTYPE html> <html lang="en"> <head> <meta charset="UTF-8"> <title>Title</title> </head> <body> <audio src="" autoplay controls id="player"></audio> <button onclick="start_reco()">录制消息</button> <br> <button onclick="stop_reco()">发送语音消息</button> </body> <script src="/static/Recorder.js"></script> <script type="application/javascript"> var serv = "http://127.0.0.1:8008"; var ws_serv = "ws://127.0.0.1:8006/ws"; var get_music = serv + "/get_audio/"; var ws = new WebSocket(ws_serv); ws.onmessage = function (data) { document.getElementById("player").src = get_music + data.data }; var reco = null; var audio_context = new AudioContext(); navigator.getUserMedia = (navigator.getUserMedia || navigator.webkitGetUserMedia || navigator.mozGetUserMedia || navigator.msGetUserMedia); navigator.getUserMedia({audio: true}, create_stream, function (err) { console.log(err) }); function create_stream(user_media) { var stream_input = audio_context.createMediaStreamSource(user_media); reco = new Recorder(stream_input); } //录制消息 function start_reco() { reco.record(); } //先停止录制，再获取音频 function stop_reco() { reco.stop(); get_audio(); reco.clear(); } //获取音频，发送音频 function get_audio() { reco.exportWAV(function (wav_file) { // wav_file = Blob对象 ws.send(wav_file); }) } </script> </html>