Python sample speech_recognition_with_push_stream blocks long time

Describe the bug Python speech sample function speech_recognition_with_push_stream contains line speech_recognizer.stop_continuous_recognition(), which blocks for 10s. I guess that 10s (10 000ms) period could be a timeout, which probably expires because speech recognizer doesn’t detect, that stream has been already closed in previous step and method is blocking and waiting until stream is closed. https://github.com/Azure-Samples/cognitive-services-speech-sdk/blob/ac8c9899c4c741a836fbcd01d66b9f4fd1d06cec/samples/python/console/speech_sample.py#L453-L457 I found out and tried these scenarios:

• when I add some debug prints

stream.close()
print("Before...")
start = time.time()
speech_recognizer.stop_continuous_recognition()
print(f"After {time.time() - start}...")

output is

Before...
RECOGNIZED: SpeechRecognitionEventArgs(session_id=<session_id>, result=SpeechRecognitionResult(result_id=<result_id>, text="", reason=ResultReason.RecognizedSpeech))
After 10.02142333984375...

• when I add time.sleep(1) after closing the stream, it is without 10s delay

stream.close()
time.sleep(1)
print("Before...")
start = time.time()
speech_recognizer.stop_continuous_recognition()
print(f"After {time.time() - start}...")

RECOGNIZED: SpeechRecognitionEventArgs(session_id=<session_id>, result=SpeechRecognitionResult(result_id=<result_id>, text="", reason=ResultReason.RecognizedSpeech))
CANCELED SpeechRecognitionCanceledEventArgs(session_id=<session_id>, result=SpeechRecognitionResult(result_id=<result_id>, text="", reason=ResultReason.Canceled))
SESSION STOPPED SessionEventArgs(session_id=<session_id>)
Before...
After 0.0006592273712158203...

• when I replace speech_recognizer.stop_continuous_recognition() by async version, everything works fine without 10s delay

stream.close()
print("Before...")
start = time.time()
speech_recognizer.stop_continuous_recognition_async().get()
print(f"After {time.time() - start}...")

Before...
RECOGNIZED: SpeechRecognitionEventArgs(session_id=<session_id>, result=SpeechRecognitionResult(result_id=<result_id>, text="", reason=ResultReason.RecognizedSpeech))
SESSION STOPPED SessionEventArgs(session_id=<session_id>)
After 0.598764181137085...

To Reproduce Steps to reproduce the behavior:

1. Run function speech_recognition_with_push_stream https://github.com/Azure-Samples/cognitive-services-speech-sdk/blob/ac8c9899c4c741a836fbcd01d66b9f4fd1d06cec/samples/python/console/speech_sample.py#L417
2. It takes 10 000 ms to stop execution of the stop_continuous_recognition method

Expected behavior stop_continuous_recognition should not block so long.

Version of the Cognitive Services Speech SDK azure-cognitiveservices-speech version 1.18.0

Platform, Operating System, and Programming Language

• OS: Ubuntu 18.04
• Hardware - x64
• Programming language: Python

[pankopon]

Fixed in the Speech SDK release 1.21.0 which is available now.

[pankopon]

Internal work item ref. 3712714.