For the AIND Project 4 we studied the use of Hidden Markov Models in solving the problem of training a models based on x and y coordinates sampled from video frames that can then match features from new unseen video frames of similar sign language.
We also learnt about ngrams from Natural Language Processing and attempted to apply the ngram probabilities of our detected sentences to improve the Word Error Rate (WER).
$ jupyter notebook asl_recognizer.ipynb