<center>![Resources #52.png](https://steemitimages.com/DQmPdJxQhSXraqJcf5DW9a2YF7hE4jN7ewShLhoDW3y2AC3/Resources%20%2352.png)</center>
___
Rachana Solanki and Stavan Himal created, as part of their final semester project at Ahmedabad University, a hand glove that converts American Sign Language to speech. Himal posted about this on his _Medium_ profile:

_" The glove works on the principles of Machine Learning Algorithm that identifies the gestures regardless of different hand sizes."_ [[source](https://medium.com/mlreview/case-study-sign-to-speech-converter-facilitated-wireless-communication-358b3914d398)]

What's even cooler is the glove is equipped with bluetooth so that two people could converse through sign language remotely within the range of 50 meters. 

Hmm, the first thing that comes to mind is how secure this connection is because we know that the Internet of Things is more often than not the internet of vulnerable/hackable things due to the lack of poorly written security protocols. Anyways, not to diverge...

Himal writes about how it all started and what their major challenges were when initiating the project. He also presents the outline of the entire process, along with a block diagram to show the architecture of the glove. 

He doesn't go too much into the machine learning/deep learning aspects of it, but he is careful enough to point out that they used a K-Nearest Neighbors (classifier) to train on the data (supervised learning):

_"As our primary focus was to deploy simple English Alphabets, we took gesture inputs of 20–25 people and recorded all the alphabetical gesture instances which were about 33,000 and we used 66.66% of the total instances as our Training Data Set and rest of the instances as our Testing Data Set."_ [[source](https://medium.com/mlreview/case-study-sign-to-speech-converter-facilitated-wireless-communication-358b3914d398)]

Their system yielded 96.9% accuracy and it's interesting that KNN worked best despite the fact that they tried other algorithms such as _ZeroR_ and a neural network. 

I'd encourage those interested to read the full post below. This type of project is laudable and such efforts are worth replicating!

<center>[Sign to Speech Converter - [Deep Learning]](https://medium.com/mlreview/case-study-sign-to-speech-converter-facilitated-wireless-communication-358b3914d398)</center>
___
### <center>To stay in touch with me, follow @cristi</center>   
___

[Cristi Vlad](http://cristivlad.com) Self-Experimenter and Author