突发奇想小程序尝试着集成语音识别功能,对号入座般首选方案为百度语音识别nodejs REST API。

其流程如图所示:

微信小程序使用百度语音识别成功后,小程序端cancel了请求_第1张图片

小程序端录制语音代码 如下所示:



  
    
      
        
      
    
  

    
  

//index.js

//录制音频参数
const options = {
  duration: 60000,
  sampleRate: 44100,
  numberOfChannels: 1,
  encodeBitRate: 192000,
  format: 'mp3',
  frameSize: 50
};
//语音录制管理器
const recorderManager = wx.getRecorderManager();

/**
   * 开始录制语音
   *
   */
  startRecordAudio:function()
  {

    recorderManager.onStart(() => {
      this.setData({ audioState: "语音录制中.." });
    });
    recorderManager.onResume(() => {
      console.log('recorder resume')
    });
    recorderManager.onPause(() => {
      this.setData({ audioState: "语音暂停录制" });
    });
    recorderManager.onStop((res) => {
      console.log('recorder stop', res);
      if (res.duration < 1000) {

        this.setData({ audioState: "开始录制" });

        wx.showToast({
          title: '录音不能小于1秒钟',
        });

        return;
      };

      var that = this; 
      wx.saveFile({
        tempFilePath: res.tempFilePath,
        success: function (res) {
          var savedFilePath = res.savedFilePath;
          that.setData({ tempAudioFilePath: savedFilePath });
          that.uploadRecordAudio();
        }
      })
    });
    recorderManager.onFrameRecorded((res) => {
      const { frameBuffer } = res
      console.log('frameBuffer.byteLength', frameBuffer.byteLength)
    });
    recorderManager.onError((res) => {
      const { error } = res;
      console.log("record error is " + error);
    });

    recorderManager.start(options);
  },

  /**
   * 结束录制语音
   *
   */
  stopRecordAudio: function () {
    this.setData({ audioState: "语音录制完成,上传中.." });
    recorderManager.stop();
  },

  /**
     * 上传录制语音
     *
     */
  uploadRecordAudio: function () {

    wx.showLoading({
      title: '音频上传中..',
    })

    this.setData({ audioState: this.data.tempAudioFilePath });
    var that = this;
    const uploadTask = wx.uploadFile({
      url: 'http://10.129.83.13:3000/aip', 
      // url: 'http://aip.butterfly.mopaasapp.com/aip',
      filePath: that.data.tempAudioFilePath,
      name: 'file',
      header: {
        'content-type': 'multipart/form-data'
      },
      formData: {
        'user': 'test'
      },
      success: function (res) {
        //do something
        console.log("res" + JSON.stringify(res));
        wx.hideLoading();
        that.echoRecordAudioContent(res.data);
      }
    });

    uploadTask.onProgressUpdate((res) => {

      this.setData({ audioState: "语音录制完成,上传中" + res.progress+".." });
      console.log('上传进度', res.progress)
      console.log('已经上传的数据长度', res.totalBytesSent)
      console.log('预期需要上传的数据总长度', res.totalBytesExpectedToSend)
    });

    // uploadTask.abort() // 取消上传任务

  },

  /**
     * 回显录制语音转化成的文字
     *
     */
    echoRecordAudioContent: function (obj) {
      console.log("文字回显" + JSON.stringify(obj));

        var title = obj.result[0];
        wx.showToast({
          title: obj.result,
        })
  }

后端代码实现采用的Nodejs+express解决方案(因为涉及到音频格式的转码 需要在本地部署ffmpeg进行音频格式的转化)

aip.js代码如下

var express = require('express');
var router = express.Router();
var fs = require('fs');
var formidable = require("formidable");
var ffmpeg = require('fluent-ffmpeg');  

var AipSpeechClient = require("baidu-aip-sdk").speech;

// 设置APPID/AK/SK
var APP_ID = "******************";
var API_KEY = "******************";
var SECRET_KEY = "******************";
// 新建一个对象,建议只保存一个对象调用服务接口
var client = new AipSpeechClient(APP_ID, API_KEY, SECRET_KEY);

router.post('/', function(req, res, next) {

    var tmp_path = null;
    var target_path = null;

    var form = new formidable.IncomingForm();
    form.uploadDir = './uploads/tmp/';
    form.maxFieldsSize = 20 * 1024 * 1024;
    form.timeout =
    form.parse(req,function(error,fields,files){

        if(!error)
        {
            tmp_path = files.file.path;
            // 指定文件上传后的目录 - 示例为"audio"目录。
            target_path = './uploads/audio/' + files.file.name;
        }
        else
        {
            res.status(200).send("{'err_msg':'语音上传失败'}");
        }
    });
    form.on('error', function(error) {
        console.log('********************error'+error);
    });
    form.on('aborted', function() {
    });
    form.on('end', function() {

        if(fs.existsSync(tmp_path))
        {
            // 移动文件
            fs.rename(tmp_path, target_path, function(rename_err) {
                if (rename_err) throw rename_err;

                //使用ffmpeg 进行音频转换
                convertAudioType(target_path).then((data) => {
                    if(data != undefined)
                    {
                        res.send(data);
                    }
                    else
                    {
                        res.status(200).send("{'err_msg':'语音上传失败'}");
                    }
                });
            });
        }

    });
});

//使用ffmpeg 进行音频转换 已经知道上传格式为aac
function convertAudioType(path)
{
    return new Promise((resolve,reject) => {

        //文件名称
        var filename = path.substr(path.lastIndexOf('/')+1);
        filename = filename.slice(0,filename.lastIndexOf('.'));

        //输出文件路径
        var outputPath = './assets/convert/audio/'+filename+'.wav';

        const extractAudio = require('ffmpeg-extract-audio');
            extractAudio({
                input: path,
                output: outputPath,
                format:'wav',
                transform: (cmd) => {
                cmd.audioFilters([
                    {
                        filter: 'volume',
                        options: '0.5'
                    }
                ]);
                cmd.on('error', function(err) {
                    if(err) return reject({"err_msg":"语音格式转换失败"});
                });
                cmd.on('end', function() {

                    var obj = convertToText(path).then((data) => {
                        console.log("******************"+JSON.stringify(data));
                        return JSON.stringify(data);
                    }).catch((error) => {

                    });

                    return resolve(obj);
                });
            }
        });

    });

}

//使用百度语音识别
function convertToText(path)
{
    return new Promise((resolve,reject) => {

        // setTimeout(function(){return resolve({"err_msg":"语音识别失败"});},1000);
        var voice = fs.readFileSync(path);
        var voiceBuffer = new Buffer(voice);

        // 识别本地文件
        client.recognize(voiceBuffer, 'wav', 16000).then(function (result) {
            var data = result;
            if(data.err_no == 0)
            {
                //{"corpus_no":"6543498202833665207","err_msg":"success.","err_no":0,"result":["欢迎使用百度语音合成,"],"sn":"321835246031523526898"}
                return resolve({"err_msg":"success","result":data.result[0]}); //小程序端直接cancel了请求
            }
            else
            {
                //{"err_msg":"speech quality error.","err_no":3301,"sn":"725439393781523526861"}
                return resolve({"err_msg":"语音识别失败"}); //小程序端可以接收到 response
            }
        }, function(err)
        {
            console.log(err);
            if(err) return reject({"err_msg":"语音识别失败"});
        });
    });

}

module.exports = router;

问题来了,红色标注如下图:

微信小程序使用百度语音识别成功后,小程序端cancel了请求_第2张图片

问题在持续跟踪中,也给百度反馈了,等待百度给出合理的解释。。。