Two Hidden Layer Multiconnected Boltzmann Machine for Controlling Moving Objects

There are various areas of industry where the remote control of machines significantly increases productivity. In terms of household this is extremely important for elderly and disabled people. Voice instructions are the most natural way of communication between employees from different hierarchical structures. The development of contemporary hardware allowing such relationships to build span between man and machine. Contact with the managed object of spoken language makes control very efficiently. Moving objects can be successfully controlled by voice commands having in mind that the voice control assures hands free and fast communication between human and controlled object. The present paper deals with the remote control of moving objects. Voice control is used for managing robots, drones, wheelchairs etc. A new stochastic classifier has been obtained for this purpose. A successful classification by a new two hidden layer Boltzmann machine has been realized. The process of deep machine learning has been studied with respect to the mean field approximation problem. An improved fixed-point iteration algorithm is used to accelerate the rate of convergence. An algorithm for training the classifier has been written explicitly in pseudocode. Real life tests are discussed.