demus/openhab-addons
1
0
Fork 0
mirror of https://github.com/openhab/openhab-addons.git synced 2025-01-10 15:11:59 +01:00
Code Issues Packages Projects Releases Wiki Activity

[googlestt] Fix drop bytes (#14649)

Browse Source 
* [googlestt] Fix drop bytes
* fix unhandled cancelation error when using single utterance mode

Signed-off-by: Miguel Álvarez <miguelwork92@gmail.com>
This commit is contained in:
GiviMAD 2023-03-23 01:57:00 -07:00 committed by Jacob Laursen
parent 3bab21f242
commit 555210cb23
1 changed files with 22 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

34
bundles/org.openhab.voice.googlestt/src/main/java/org/openhab/voice/googlestt/internal/GoogleSTTService.java
View File

@ -255,14 +255,20 @@ public class GoogleSTTService implements STTService {
long startTime = System.currentTimeMillis();
long maxTranscriptionMillis = (config.maxTranscriptionSeconds * 1000L);
long maxSilenceMillis = (config.maxSilenceSeconds * 1000L);
int readBytes = 6400;
while (!aborted.get()) {
byte[] data = new byte[readBytes];
int dataN = audioStream.read(data);
final int bufferSize = 6400;
int numBytesRead;
int remaining = bufferSize;
byte[] audioBuffer = new byte[bufferSize];
while (!aborted.get() && !responseObserver.isDone()) {
numBytesRead = audioStream.read(audioBuffer, bufferSize - remaining, remaining);
if (aborted.get()) {
logger.debug("Stops listening, aborted");
break;
}
if (numBytesRead == -1) {
logger.debug("End of stream");
break;
}
if (isExpiredInterval(maxTranscriptionMillis, startTime)) {
logger.debug("Stops listening, max transcription time reached");
break;
@ -272,18 +278,17 @@ public class GoogleSTTService implements STTService {
logger.debug("Stops listening, max silence time reached");
break;
}
if (dataN != readBytes) {
try {
Thread.sleep(100);
} catch (InterruptedException e) {
}
if (numBytesRead != remaining) {
remaining = remaining - numBytesRead;
continue;
}
remaining = bufferSize;
StreamingRecognizeRequest dataRequest = StreamingRecognizeRequest.newBuilder()
.setAudioContent(ByteString.copyFrom(data)).build();
logger.debug("Sending audio data {}", dataN);
.setAudioContent(ByteString.copyFrom(audioBuffer)).build();
logger.debug("Sending audio data {}", bufferSize);
clientStream.send(dataRequest);
}
audioStream.close();
}
private void sendStreamConfig(ClientStream<StreamingRecognizeRequest> clientStream,
@ -335,6 +340,7 @@ public class GoogleSTTService implements STTService {
private float confidenceSum = 0;
private int responseCount = 0;
private long lastInputTime = 0;
private boolean done = false;
public TranscriptionListener(STTListener sttListener, GoogleSTTConfiguration config, AtomicBoolean aborted) {
this.sttListener = sttListener;
@ -374,7 +380,7 @@ public class GoogleSTTService implements STTService {
responseCount++;
// when in single utterance mode we can just get one final result so complete
if (config.singleUtteranceMode) {
onComplete();
done = true;
}
}
});
@ -411,6 +417,10 @@ public class GoogleSTTService implements STTService {
}
}
public boolean isDone() {
return done;
}
public long getLastInputTime() {
return lastInputTime;
}