IBM的语音识别（IBM speech to text 语言转换成文字）

1.登陆网址https://www.ibm.com/watson/developercloud/speech-to-text.html并注册

2.打开网址https://console.ng.bluemix.net/catalog/?category=watson，点击界面左侧菜单中的Watson,并选择Speech ToText

3.点击界面最下方的创建按钮

4.点击新建凭证

5.输入名称，点击添加

6.点击查看凭证，即可看到账号、密码和请求地址（凭证创建完成后，不一定能马上使用，有时需要过一段时间，具体时间不确定，我当天申请的当天测试一直连接不上，两天后再测的时候就可以了）

7.创建一个工程，将speech-android-wrapper作为library导入，并添加到工程中，build.gradle中的SDK版本号不要超过23，否则会报org.apache.http找不到，因为该api在23以后更改了
（例子代码github地址：https://github.com/watson-developer-cloud/speech-android-sdk）

8.新建的项目中添加依赖

9.主界面布局

10.授权

    if (initSTT() == false) {
            displayResult("Error: no authentication credentials/token available, please enter your authentication information");
            return;
        }

    private boolean initSTT() {
        // DISCLAIMER: please enter your credentials or token factory in the lines below
        String username = "173e235d-5d3d-453e-9d97-0b0f77bdac19";                         //账号
        String password = "vjpqIUx0Ss8x";						  //密码
        String serviceURL = "wss://stream.watsonplatform.net/speech-to-text/api";         //服务器地址

        SpeechConfiguration sConfig = new SpeechConfiguration(SpeechConfiguration.AUDIO_FORMAT_OGGOPUS);
        sConfig.learningOptOut = false; // Change to true to opt-out

        SpeechToText.sharedInstance().initWithContext(this.getHost(serviceURL), context, sConfig);   	//设置服务器地址
        SpeechToText.sharedInstance().setCredentials(username, password);				//设置账号和密码

        SpeechToText.sharedInstance().setModel("en-US_BroadbandModel");					//设置所采集的语言
        SpeechToText.sharedInstance().setDelegate(this);						//设置监听
        return true;
    }

11.检测服务是否连接

if (jsonModels == null) {
            jsonModels = new STTCommands().doInBackground();
            if (jsonModels == null) {
                displayResult("Please, check internet connection.");  					//检测服务是否连接
                return;
            }
        }

}

12.重写监听中的方法

public class MainActivity extends Activity implements ISpeechDelegate    //activity实现了该监听，并复写监听中的方法

 @Override
    public void onOpen() {					//认证开始和服务器连接上
        Log.d(TAG, "onOpen");
        mState = ConnectionState.CONNECTED;
    }

    @Override
    public void onError(String error) {						//连接出错
        Log.e(TAG, "onError...................." + error);
        mState = ConnectionState.IDLE;
    }

    @Override
    public void onClose(int code, String reason, boolean remote) {			//连接关闭
        Log.d(TAG, "onClose, code: " + code + " reason: " + reason);
        mState = ConnectionState.IDLE;
    }

    @Override
    public void onMessage(String message) {						//获取到识别到的信息并显示到界面上
        Log.d(TAG, "onMessage, message: " + message);
        try {
            JSONObject jObj = new JSONObject(message);
            // state message
            if (jObj.has("state")) {
                Log.d(TAG, "Status message: " + jObj.getString("state"));
            }
            // results message
            else if (jObj.has("results")) {
                //if has result
                Log.d(TAG, "Results message: ");
                JSONArray jArr = jObj.getJSONArray("results");
                for (int i = 0; i < jArr.length(); i++) {
                    JSONObject obj = jArr.getJSONObject(i);
                    JSONArray jArr1 = obj.getJSONArray("alternatives");
                    String str = jArr1.getJSONObject(0).getString("transcript");
                    // remove whitespaces if the language requires it
//                    String model = "en-US_MichaelVoice";
                    String model = "en-US_BroadbandModel";
                    if (model.startsWith("ja-JP") || model.startsWith("zh-CN")) {
                        str = str.replaceAll("\\s+", "");
                    }
                    String strFormatted = Character.toUpperCase(str.charAt(0)) + str.substring(1);
                    if (obj.getString("final").equals("true")) {
                        String stopMarker = (model.startsWith("ja-JP") || model.startsWith("zh-CN")) ? "。" : ". ";
                        mRecognitionResults += strFormatted.substring(0, strFormatted.length() - 1) + stopMarker;

                        displayResult(mRecognitionResults);
                    } else {
                        displayResult(mRecognitionResults + strFormatted);
                    }
                    break;
                }
            } else {
                displayResult("unexpected data coming from stt server: \n" + message);
            }

        } catch (JSONException e) {
            Log.e(TAG, "Error parsing JSON");
            e.printStackTrace();
        }
    }

点击record，说Test，点击stop，然后数据返回并设置到text上

13.点击Record按钮时进行的操作

 btRecord.setOnClickListener(new View.OnClickListener() {
            @Override
            public void onClick(View v) {
                if (mState == ConnectionState.IDLE) {
                    mState = ConnectionState.CONNECTING;
                    Log.d(TAG, "onClickRecord: IDLE -> CONNECTING");
                    mRecognitionResults = "";
                    displayResult(mRecognitionResults);
                    SpeechToText.sharedInstance().setModel("en-US_BroadbandModel");			//设置识别的语言是英语
                    // start recognition
                    new AsyncTask<Void, Void, Void>() {
                        @Override
                        protected Void doInBackground(Void... none) {
                            SpeechToText.sharedInstance().recognize();					//开始识别
                            return null;
                        }
                    }.execute();
                    btRecord.setText("Connecting...");
                }
            }
        });

14.停止识别

 btStop.setOnClickListener(new View.OnClickListener() {
            @Override
            public void onClick(View v) {
                if (mState == ConnectionState.CONNECTED) {
                    mState = ConnectionState.IDLE;
                    Log.d(TAG, "onClickRecord: CONNECTED -> IDLE");
                    SpeechToText.sharedInstance().stopRecognition();                                   //停止识别
                    btRecord.setText("Record");
                }
            }
        });

注：1.build.gradle中的compile和build版本不要超过21

2.服务器是国外的，需要连接vpn

3.在回调方法中，是子线程，不要进行更改界面的操作，如果需要更新界面，需先运行在主线程

参考资料：

https://www.ibm.com/watson/developercloud/speech-to-text.html