Generating Handwriting from Multichannel Electromyographic Activity
Handwriting is an advanced motor skill and one of the key developments in human culture. Here we show that handwriting can be decoded—offline and online—from electromyographic (EMG) signals recorded from multiple hand and forearm muscles. We convert EMGs into continuous handwriting traces and into discretely decoded font characters. For this purpose, we use Wiener and Kalman filters, and machine learning algorithms. Our approach is applicable to clinical neural prostheses for restoration of dexterous hand movements, and to medical diagnostics of neural disorders that affect handwriting. We also propose that handwriting could be decoded from cortical activity, such as the activity recorded with electrocorticography (ECoG).