Keystroke dynamics for arbitrary length text?

Question

Algorithms

Interested in such a thing as biometric authentication. The first thing that caught my attention was Keystroke_dynamics
. I started to study this direction, but faced the problem that most of the research was aimed at analyzing the input of passwords + logins, which are constants in themselves (i.e. it is not difficult to train the machine based on the time that the key was pressed Dwell time, and the transition time between the keys Flight time). I have not found any research in the field of text analysis of arbitrary length.
Still wondering what learning algorithms can be studied related to this topic? For as I understand it, there is nowhere without machine learning :)

Answer 1

Text of arbitrary length, apparently, break into fragments and do a statistical analysis.
As far as machine learning is concerned, features are distinguished and remembered (a dictionary is created). This method is used for image recognition.
And here is an example of how smartphone accelerometer signals are processed to find out the position of a person (standing, lying, walking, walking up or down stairs). The signal is divided into fragments (buffers), then there is a spectral analysis, the features of the spectrum are remembered (the neural network is trained).