Issue on continuous Voice Recognition , duplicating voice input and sometimes it splitting single voice to multiple text

Question

Issue on continuous Voice Recognition , duplicating voice input and sometimes it splitting single voice to multiple text

Anandu K B 0

Could you please provide a code example for implementing continuous voice recognition that can effectively avoid or reduce background noise?

In my current implementation, I am facing the following issues:

Background Noise Interference: The recognition is frequently triggered by surrounding noises, which results in incorrect or unwanted transcriptions.

Sentence Splitting Issue: When speaking into the microphone, sometimes a single sentence is being recognized as two separate sentences. This splitting affects the overall transcription accuracy and user experience.

My goal is to have continuous speech recognition that:

Ignores/minimizes background noise.

Correctly recognizes spoken sentences without unintended splitting.

Please share any code samples, configurations, or best practices to address these issues.Could you please provide a code example for implementing continuous voice recognition that can effectively avoid or reduce background noise?

In my current implementation, I am facing the following issues:

Background Noise Interference:
The recognition is frequently triggered by surrounding noises, which results in incorrect or unwanted transcriptions.

Sentence Splitting Issue:
When speaking into the microphone, sometimes a single sentence is being recognized as two separate sentences. This splitting affects the overall transcription accuracy and user experience.

My goal is to have continuous speech recognition that:

Ignores/minimizes background noise.

Correctly recognizes spoken sentences without unintended splitting.

Please share any code samples, configurations, or best practices to address these issues.

1 answer

Your answer

Answer 1

hi Anandu, merci for posting this on the QnA )) ok lets get u a clean continuous speech setup that keeps the mic open, calms the background buzz, and stops cutting one sentence into two. azure can do this nicely if u tune three things at once recognition mode, silence timeouts, and the input audio path https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-recognize-speech

for microsoft setup i would start with continuous recognition and set explicit silence knobs. initial silence says how long we wait for the first words. end silence says when we decide the sentence ended. segmentation silence says how long a pause can live inside one sentence before we chop it. all three exist as properties in the sdk and they work in continuous mode. this helps a ton with the random split problem. take a look here and keep it handy https://learn.microsoft.com/en-us/javascript/api/microsoft-cognitiveservices-speech-sdk/propertyid and here for continuous start and events https://learn.microsoft.com/en-us/dotnet/api/microsoft.cognitiveservices.speech.speechrecognizer.startcontinuousrecognitionasync

now the noise side. the speech sdk ships with microsoft audio stack that brings noise suppression echo kill dereverb and beamforming when supported. turn it on and let the stack do the heavy lifting. this reduces false triggers from room noise and keyboard taps. read the short intro here https://learn.microsoft.com/en-us/azure/ai-services/speech-service/audio-processing-overview and the how to for the audio stack here https://docs.azure.cn/en-us/ai-services/speech-service/audio-processing-speech-sdk as well check this guidance in voice live where noise suppression is called out directly https://learn.microsoft.com/en-us/azure/ai-services/speech-service/voice-live-how-to

Alex

and "yes" if you would follow me at Q&A - personaly thx.
P.S. If my answer help to you, please Accept my answer

https://ctrlaltdel.blog/

Share via

Issue on continuous Voice Recognition , duplicating voice input and sometimes it splitting single voice to multiple text

1 answer

Your answer