This article shows the acquisition of electrical signals from the nerves of the throat and oral cords, transforming them into voice signals using advanced techniques in digital signal processing and artificial intelligence, based on the extraction of patterns based on compressive sensing, Entropy, discrete Wavelet transform and a classifier based on vector support machines of least squares, once the system has been calibrated, the results showed that 95% + - 0.34 of data were correctly classified. The developed system was used in 500 signals and is based on an Open Source programming language implemented in an embedded system. Finally, it was shown that it is possible to use compressive sensing to extract subvocal speech patterns.