Man solves the problems of recognition every day, every minute, and may be every second. These problems include picking out your close ones or friends in a crowd of people, identifying the voice of a caller, recognizing a brand or product when shopping, remembering a location you visited before, etc.
Beyond these cognitive events occurring in our personal every day life, there are also recognition problems which we solve at work, in an industrial or enterprise environment. For example, operating a process control unit, an instrument or other equipment. In such a situation, a human operator controls a set of parameters with the objective to maintain them within a certain range of values. Depending on the combination of these values, one process can be classified as Normal, Alert or Alarm, or another process can recognize a product with Grade 1or 2, or a product model A, B, C or D.
UAB Neurotechnologijos offers an NТ Industrial Automation Solution designed to solve pattern recognition problems in images, live and recorded video stream, audio signals, vibrations and other free-form signals. An NT Industrial Automation system consists of an NT Adaptive Controller and one or more software core module running on a server.
The NT Adaptive controllers come in two flavors to support either analog signal or video input channels. They all have in common a unique neural network hardware capable of real-time learning and recognition, novelty and anomaly detection. This neural network is called a NeuroMem neural network. It is assembled from a bank of identical NeuroMem neuromorphic chips and the number of these chips defines the network capacity. Unlike other neural network hardware, a NeuroMem network is digital, easy to train and query, low-power, scalable and it is not a black box. Indeed, the models retained by the neurons can be traced and their behavior can be explained! For more information about the NeuroMem™ neural networks, please refer to our NeuroMem™ technology page.