Week 6

 This week I looked at "Mind the Uuh", an experimental training tool that aims to improve users' public speaking skills. It was developed by Benedikt Groß, Maik Groß, and Thibault Durand.  This project was created for Google’s TensorFlow Microcontroller Challenge, a competition from 2021. The device constantly monitors the sound of speech and alerts users upon the use of filler words such as "uhh". 

The prototype for this project was purposefully made as uncomplicated as possible. It consists of a bell, a volume turner button that determines how loud the bell sounds, a counter for the amount of "uhh's" uttered, and a reset button. The product design purposefully makes allusions to vintage alarm clocks in order to intuitively explain the nature of Mind the "Uuh". It utilises a fear tactic that aims to inspire confidence in its users, avoiding long pauses and nonsense words. 

The small gadget utilises a custom trained machine learning model that is trained to identify the "uuh"s, It  studied over 1500 samples ranging from lengths of 300 milliseconds to 1 second. 

This is currently just a proof of concept model, so it can pick up on easily recognisable "uuh's", but will miss small or brief "uuh's". I find this incredibly impressive regardless. Nothing is transferred to the cloud; all speech data is handled natively on the small device.

Here is the layout of the Arduino held inside the casing.

I am inspired by the combination of new technology held within an case in the style of an older gadget.