Hi @ShivaRagu ,
I believe it is possible, even if not so straightforward. You might refer to the general solution provided by ST, that however runs on STM32H7, while the Sensortile.box supports STM32L4: I'm referring to the STEVAL-VOICE-UI Voice User Interface reference design kit.
Another possible solution, directly involving the Sensortile.box tool, can be use the Qeexo ML tool: this allows you to run many kind of neural networks on the audio acquisition from the analog MEMS microphone on the Sensortile.box tool.
-Eleon