[Python] is there any method to capture and transcribe the audio of a person in front of...

I am working on a RoboDog project. I have developed the realtime audio and video capturing pipeline. The audio pipeline is used to talk with robodog (listen audio, transcribe it and send it to robodog). While the video pipeline is used for face and Qr code detection. The main hurdle i am facing is that the audio can contain a lot of background noise of other persons (e.g in a conference) and other devices like loud speakers, fans etc. I want a pipeline for speech recognition, that recognize and transcribe the speech of the person in front of robodog camera only. Mean we have to implement some method to filter out and transcribe the audio of person in front of camera.

Here are some important considerations:

• The other persons (in background may speaking loudly and their audio can dominate). we have to handle this situation.
• For now we have limited known people for person recognition using roodog camera.
• I have tried some audio visual speech recognition models, but their performance is not satisfactory.
• I am using Python for development

Please suggest me some approach to solve this problem.

Continue reading...