Entropy to Release Geometric Distance
JAPAN - Entropy Software Laboratory, a voice and image recognition technology company, is releasing a new solutuion to a conventional "wobble" problem. In conventional Euclid distance and similarity, calculation is done on the standard and input patterns for relevant components, without taking the mutual relationship with surrounding components into consideration. Therefore the conventional method is susceptible to "wobble" such as noise. Entropy created a new similarity scale called Geometric Distance and developed the pattern matching method which can precisely detect the figure of distance between patterns. In the case of Geometric Distance, calculation is done so that it meets human sensitivities considering the mutual relationship with surrounding components. If the conventional Euclid distance and similarity methods are replaced with Entropy's technology of Geometric Distance, the performance of voice and image recognition improves.