Merge pull request #1068 from TakahiroSakuda/fix_stt_audio_metrics

Fix AudioMetrics bug (Speech To Text)

Author: lpatino10

speech-to-text/src/main/java/com/ibm/watson/speech_to_text/v1/websocket/SpeechToTextWebSocketListener.java

@@ -54,6 +54,7 @@ public final class SpeechToTextWebSocketListener extends WebSocketListener {
   private static final String ERROR = "error";
   private static final String RESULTS = "results";
   private static final String SPEAKER_LABELS = "speaker_labels";
+  private static final String AUDIO_METRICS = "audio_metrics";
   private static final String CUSTOMIZATION_ID = "customization_id";
   private static final String LANGUAGE_CUSTOMIZATION_ID = "language_customization_id";
   private static final String ACOUSTIC_CUSTOMIZATION_ID = "acoustic_customization_id";
@@ -129,7 +130,7 @@ public void onMessage(WebSocket webSocket, String message) {
         // notify that the service timeouts because of inactivity
         callback.onInactivityTimeout(new RuntimeException(error));
       }
-    } else if (json.has(RESULTS) || json.has(SPEAKER_LABELS)) {
+    } else if (json.has(RESULTS) || json.has(SPEAKER_LABELS) || json.has(AUDIO_METRICS)) {
       callback.onTranscription(GSON.fromJson(message, SpeechRecognitionResults.class));
 
     } else if (json.has(STATE)) {

speech-to-text/src/test/java/com/ibm/watson/speech_to_text/v1/SpeechToTextIT.java

@@ -103,7 +103,8 @@ public class SpeechToTextIT extends WatsonServiceTest {
 
   private CountDownLatch lock = new CountDownLatch(1);
   private SpeechToText service;
-  private SpeechRecognitionResults asyncResults;
+  private SpeechRecognitionResults asyncTranscriptionResults;
+  private SpeechRecognitionResults asyncAudioMetricsResults;
   private Boolean inactivityTimeoutOccurred;
   private String customizationId;
   private String acousticCustomizationId;
@@ -221,11 +222,13 @@ public void testRecognizeFileStringRecognizeOptions() throws FileNotFoundExcepti
         .model(EN_BROADBAND16K)
         .contentType(contentType)
         .profanityFilter(false)
+        .audioMetrics(true)
         .build();
     SpeechRecognitionResults results = service.recognize(options).execute().getResult();
     assertNotNull(results.getResults().get(0).getAlternatives().get(0).getTranscript());
     assertNotNull(results.getResults().get(0).getAlternatives().get(0).getTimestamps());
     assertNotNull(results.getResults().get(0).getAlternatives().get(0).getWordConfidence());
+    assertNotNull(results.getAudioMetrics());
   }
 
   /**
@@ -286,6 +289,7 @@ public void testRecognizeWebSocket() throws FileNotFoundException, InterruptedEx
         .interimResults(true)
         .processingMetrics(true)
         .processingMetricsInterval(0.2f)
+        .audioMetrics(true)
         .build();
 
     service.recognizeUsingWebSocket(options, new BaseRecognizeCallback() {
@@ -314,22 +318,29 @@ public void onError(Exception e) {
 
       @Override
       public void onTranscription(SpeechRecognitionResults speechResults) {
-        if (speechResults != null && speechResults.getResults().get(0).isFinalResults()) {
-          asyncResults = speechResults;
+        if (speechResults != null) {
+          if (speechResults.getResults() != null && speechResults.getResults().get(0).isFinalResults()) {
+            asyncTranscriptionResults = speechResults;
+          }
+          if (speechResults.getAudioMetrics() != null) {
+            asyncAudioMetricsResults = speechResults;
+          }
           System.out.println(speechResults);
         }
       }
 
     });
 
     lock.await(2, TimeUnit.MINUTES);
-    assertNotNull(asyncResults);
+    assertNotNull(asyncTranscriptionResults);
+    assertNotNull(asyncAudioMetricsResults);
 
-    List<WordAlternativeResults> wordAlternatives = asyncResults.getResults().get(asyncResults.getResultIndex()
-        .intValue()).getWordAlternatives();
+    List<WordAlternativeResults> wordAlternatives = asyncTranscriptionResults.getResults()
+        .get(asyncTranscriptionResults.getResultIndex().intValue()).getWordAlternatives();
     assertTrue(wordAlternatives != null && !wordAlternatives.isEmpty());
     assertNotNull(wordAlternatives.get(0).getAlternatives());
-    assertNotNull(asyncResults.getProcessingMetrics());
+    assertNotNull(asyncTranscriptionResults.getProcessingMetrics());
+    assertNotNull(asyncAudioMetricsResults.getAudioMetrics());
   }
 
   /**