When I use play_and_detect_speech, it affects my regular voice calls.

[wangxm345566462]

Describe the bug
When I enable continuous speech recognition with the command say: detect:unimrcp:unimrcpserver-mrcp-v2-asr1 {start-input-timers=false,recognition-timeout=25000,play_and_detect_speech_close_asr=false}jt.continuous.grxml, the audio stream in my FreeSWITCH session becomes disrupted. Although the audio can be sent to MRCP for recognition, the media stream in the FreeSWITCH call experiences interruptions, loss, or poor quality, leading to an inconsistent listening experience.

To Reproduce
Steps to reproduce the behavior:

Configure FreeSWITCH with MRCP for ASR using the command say: detect:unimrcp:unimrcpserver-mrcp-v2-asr1 {start-input-timers=false,recognition-timeout=25000,play_and_detect_speech_close_asr=false}jt.continuous.grxml.
Initiate a call using Verto.
Stream audio into the call, aiming for continuous speech recognition.
Observe that the audio in the FreeSWITCH call suffers from interruptions or quality degradation.
Expected behavior
I expected the continuous speech recognition to process the audio stream without impacting the quality or consistency of the audio heard during the FreeSWITCH call. The audio should be clearly audible without interruptions or loss, while also being sent to MRCP for recognition.

Package version or git hash

Version 1.8.7 and other
Trace logs
Unfortunately, I do not have trace logs available at the moment. However, I can enable DEBUG and UUID logging in FreeSWITCH to collect detailed logs if required.

backtrace from core file
Not applicable as FreeSWITCH did not crash, but experienced audio stream issues.

Please note: Adjust the steps under "To Reproduce" to closely match your actual setup and actions that led to the problem. Additionally, providing exact package versions or git hashes, along with any trace logs or core file backtraces (if FreeSWITCH crashes), can significantly help in diagnosing the issue.