Add Custom Speech integration to SDK and sample

samples/browser/Sample.html

@@ -18,9 +18,22 @@ <h1 style="font-weight:500;">Speech Recognition</h1>
                     <h2 style="font-weight:500;">Microsoft Cognitive Services</h2>
                 </td>
             </tr>
+            <tr>
+                <td align="right">Recognition API:</td>
+                <td align="left">
+                    <select id="recognitionAPI">
+                        <option value="Bing Speech" selected="selected">Bing Speech</option>
+                        <option value="Custom Speech">Custom Speech</option>
+                    </select>
+                </td>
+            </tr>
+            <tr>
+                <td align="right"><a href="https://www.microsoft.com/cognitive-services/en-us/sign-up" target="_blank">Custom Speech Endpoint ID<br/>(leave blank for Bing Speech)</a>:</td>
+                <td><input id="endpointId" type="text" size="40" value="YOUR_CUSTOM_SPEECH_ENDPOINT_ID"></td>
+            </tr>
             <tr>
                 <td align="right"><a href="https://www.microsoft.com/cognitive-services/en-us/sign-up" target="_blank">Subscription</a>:</td>
-                <td><input id="key" type="text" size="40" value="YOUR_BING_SPEECH_API_KEY"></td>
+                <td><input id="key" type="text" size="40" value="YOUR_API_SUBSCRIPTION_KEY"></td>
             </tr>
             <tr>
                 <td align="right">Language:</td>
@@ -137,7 +150,7 @@ <h2 style="font-weight:500;">Microsoft Cognitive Services</h2>
         }
 
         // Setup the recognizer
-        function RecognizerSetup(SDK, recognitionMode, language, format, subscriptionKey) {
+        function RecognizerSetup(SDK, recognitionMode, language, format, subscriptionKey, recognitionAPI, endpointId) {
 
             switch (recognitionMode) {
                 case "Interactive" :
@@ -153,15 +166,28 @@ <h2 style="font-weight:500;">Microsoft Cognitive Services</h2>
                     recognitionMode = SDK.RecognitionMode.Interactive;
             }
 
+            switch (recognitionAPI) {
+                case "Bing Speech": 
+                    recognitionAPI = SDK.RecognitionAPI.BingSpeech;
+                    break;
+                case "Custom Speech":
+                    recognitionAPI = SDK.RecognitionAPI.CustomSpeech;
+                    break;
+                default:
+                    recognitionAPI = SDK.RecognitionAPI.BingSpeech;
+                    break;
+            }
+
             var recognizerConfig = new SDK.RecognizerConfig(
                 new SDK.SpeechConfig(
                     new SDK.Context(
                         new SDK.OS(navigator.userAgent, "Browser", null),
                         new SDK.Device("SpeechSample", "SpeechSample", "1.0.00000"))),
                 recognitionMode,
                 language, // Supported languages are specific to each recognition mode. Refer to docs.
-                format); // SDK.SpeechResultFormat.Simple (Options - Simple/Detailed)
-
+                format, // SDK.SpeechResultFormat.Simple (Options - Simple/Detailed)
+                recognitionAPI, // supports Bing Speech and Custom Speech
+                endpointId); // Custom Speech subscription key, see https://cris.ai portal
 
             var useTokenAuth = false;
 
@@ -267,10 +293,10 @@ <h2 style="font-weight:500;">Microsoft Cognitive Services</h2>
     <!-- Browser Hooks -->
     <script>
         var startBtn, stopBtn, hypothesisDiv, phraseDiv, statusDiv;
-        var key, languageOptions, formatOptions, recognitionMode, inputSource, filePicker;
+        var endpointId, key, languageOptions, formatOptions, recognitionMode, inputSource, filePicker;
         var SDK;
         var recognizer;
-        var previousSubscriptionKey;
+        var previousEndpointId, previousSubscriptionKey;
 
         document.addEventListener("DOMContentLoaded", function () {
             createBtn = document.getElementById("createBtn");
@@ -279,6 +305,7 @@ <h2 style="font-weight:500;">Microsoft Cognitive Services</h2>
             phraseDiv = document.getElementById("phraseDiv");
             hypothesisDiv = document.getElementById("hypothesisDiv");
             statusDiv = document.getElementById("statusDiv");
+            endpointId = document.getElementById("endpointId");
             key = document.getElementById("key");
             languageOptions = document.getElementById("languageOptions");
             formatOptions = document.getElementById("formatOptions");
@@ -289,16 +316,22 @@ <h2 style="font-weight:500;">Microsoft Cognitive Services</h2>
             languageOptions.addEventListener("change", Setup);
             formatOptions.addEventListener("change", Setup);
             recognitionMode.addEventListener("change", Setup);
+            recognitionAPI.addEventListener("change", Setup);
 
             startBtn.addEventListener("click", function () {
-                if (key.value == "" || key.value == "YOUR_BING_SPEECH_API_KEY") {
-                    alert("Please enter your Bing Speech subscription key!");
+                if (recognitionAPI.value === "Custom Speech" && (endpointId.value == "" || endpointId.value == "YOUR_CUSTOM_SPEECH_ENDPOINT_ID")) {
+                    alert("Please enter your Custom Speech endpoint id!");
+                    return;
+                }
+                if (key.value == "" || key.value == "YOUR_API_SUBSCRIPTION_KEY") {
+                    alert("Please enter your Bing Speech or Custom Speech API subscription key!");
                     return;
                 }
                 if (inputSource.value === "File") {
                     filePicker.click();
                 } else {
-                    if (!recognizer || previousSubscriptionKey != key.value) {
+                    if (!recognizer || previousSubscriptionKey != key.value || previousEndpointId != endpointId.value) {
+                        previousEndpointId = endpointId.value;
                         previousSubscriptionKey = key.value;
                         Setup();
                     }
@@ -311,15 +344,27 @@ <h2 style="font-weight:500;">Microsoft Cognitive Services</h2>
                 }                
             });
 
+            endpointId.addEventListener("focus", function () {
+               if (endpointId.value == "YOUR_CUSTOM_SPEECH_ENDPOINT_ID") {
+                endpointId.value = "";
+               }
+            });
+
+            endpointId.addEventListener("focusout", function () {
+               if (endpointId.value == "") {
+                endpointId.value = "YOUR_CUSTOM_SPEECH_ENDPOINT_ID";
+               }
+            });
+
             key.addEventListener("focus", function () {
-               if (key.value == "YOUR_BING_SPEECH_API_KEY") {
+               if (key.value == "YOUR_API_SUBSCRIPTION_KEY") {
                    key.value = "";
                }
             });
 
             key.addEventListener("focusout", function () {
                if (key.value == "") {
-                   key.value = "YOUR_BING_SPEECH_API_KEY";
+                   key.value = "YOUR_API_SUBSCRIPTION_KEY";
                }
             });
 
@@ -349,7 +394,7 @@ <h2 style="font-weight:500;">Microsoft Cognitive Services</h2>
             if (recognizer != null) {
                 RecognizerStop(SDK, recognizer);
             }
-            recognizer = RecognizerSetup(SDK, recognitionMode.value, languageOptions.value, SDK.SpeechResultFormat[formatOptions.value], key.value);
+            recognizer = RecognizerSetup(SDK, recognitionMode.value, languageOptions.value, SDK.SpeechResultFormat[formatOptions.value], key.value, recognitionAPI.value, endpointId.value);
         }
 
         function UpdateStatus(status) {

src/common.browser/WebsocketConnection.ts

@@ -12,6 +12,7 @@ import {
     PlatformEvent,
     Promise,
 } from "../common/Exports";
+import { RecognitionAPI } from "../sdk/speech/Exports";
 import { WebsocketMessageAdapter } from "./WebsocketMessageAdapter";
 
 export class WebsocketConnection implements IConnection {
@@ -27,6 +28,8 @@ export class WebsocketConnection implements IConnection {
         queryParameters: IStringDictionary<string>,
         headers: IStringDictionary<string>,
         messageFormatter: IWebsocketMessageFormatter,
+        recognitionAPI: RecognitionAPI,
+        endpointId: string,
         connectionId?: string) {
 
         if (!uri) {
@@ -41,6 +44,10 @@ export class WebsocketConnection implements IConnection {
 
         let queryParams = "";
         let i = 0;
+        if (recognitionAPI === RecognitionAPI.CustomSpeech) {
+            queryParams = "?cid=" + endpointId;
+            i++;
+        }
 
         if (queryParameters) {
             for (const paramName in queryParameters) {

src/sdk/speech.browser/SpeechConnectionFactory.ts

@@ -9,6 +9,7 @@ import {
     AuthInfo,
     IAuthentication,
     IConnectionFactory,
+    RecognitionAPI,
     RecognitionMode,
     RecognizerConfig,
     SpeechResultFormat,
@@ -25,6 +26,18 @@ export class SpeechConnectionFactory implements IConnectionFactory {
         authInfo: AuthInfo,
         connectionId?: string): IConnection => {
 
+        switch (config.RecognitionAPI) {
+            case RecognitionAPI.BingSpeech:
+                Storage.Local.Set("Host", "wss://speech.platform.bing.com");
+                break;
+            case RecognitionAPI.CustomSpeech:
+                Storage.Local.Set("Host", "wss://westus.stt.speech.microsoft.com");
+                break;
+            default:
+                Storage.Local.Set("Host", "wss://speech.platform.bing.com");
+                break;
+        }
+
         let endpoint = "";
         switch (config.RecognitionMode) {
             case RecognitionMode.Conversation:
@@ -51,7 +64,7 @@ export class SpeechConnectionFactory implements IConnectionFactory {
         headers[authInfo.HeaderName] = authInfo.Token;
         headers[ConnectionIdHeader] = connectionId;
 
-        return new WebsocketConnection(endpoint, queryParams, headers, new WebsocketMessageFormatter(), connectionId);
+        return new WebsocketConnection(endpoint, queryParams, headers, new WebsocketMessageFormatter(), config.RecognitionAPI, config.EndpointId, connectionId);
     }
 
     private get Host(): string {

src/sdk/speech/RecognizerConfig.ts

@@ -1,4 +1,9 @@
 
+export enum RecognitionAPI {
+    BingSpeech,
+    CustomSpeech,
+}
+
 export enum RecognitionMode {
     Interactive,
     Conversation,
@@ -16,17 +21,23 @@ export class RecognizerConfig {
     private format: SpeechResultFormat;
     private speechConfig: SpeechConfig;
     private recognitionActivityTimeout: number;
+    private recognitionAPI: RecognitionAPI;
+    private endpointId: string;
 
     constructor(
         platformConfig: SpeechConfig,
         recognitionMode: RecognitionMode = RecognitionMode.Interactive,
         language: string = "en-us",
-        format: SpeechResultFormat = SpeechResultFormat.Simple) {
+        format: SpeechResultFormat = SpeechResultFormat.Simple,
+        recognitionAPI: RecognitionAPI,
+        endpointId: string) {
         this.speechConfig = platformConfig ? platformConfig : new SpeechConfig(new Context(null, null));
         this.recognitionMode = recognitionMode;
         this.language = language;
         this.format = format;
         this.recognitionActivityTimeout = recognitionMode === RecognitionMode.Interactive ? 8000 : 25000;
+        this.recognitionAPI = recognitionAPI,
+        this.endpointId = endpointId;
     }
 
     public get RecognitionMode(): RecognitionMode {
@@ -52,6 +63,14 @@ export class RecognizerConfig {
     public get IsContinuousRecognition(): boolean {
         return this.recognitionMode !== RecognitionMode.Interactive;
     }
+
+    public get RecognitionAPI(): RecognitionAPI {
+        return this.recognitionAPI;
+    }
+
+    public get EndpointId(): string {
+        return this.endpointId;
+    }
 }
 
 // tslint:disable-next-line:max-classes-per-file