Sound recognition - advantage of 24 bit depth sound capture v 16 bit?


This is a question on the advantage of using 24 bit depth sound capture v 16 bit.

We are using an embedded device (ESP32) capturing sound using an TDK/ INVENSENSE ICS-43434 I2S microphone. It captures sound in 24 bit depth, but most firmware examples use 16 bit. We would like to minimize processing (& hence power consumption as its a battery powered device).

The question is: Is there a noticeable benefit of using 24 bit sound v 16 bit?