Method Of Speech Reconnition Using Multimodal Variational Inference With Switching State Space Models

Abstract: A method of efficently setting posterior probility parameters for a switching state space model begins by defining a window containing at least two but fewer than all of the framess. A separate posterior probability parameter is determined for each frame in the window. The window is then shifted sequentially from left to right in time so that it includes one or more subsequent frames int he sequence of frames. A separate posterior probability parameter is then determined for each frame in the shifted window. This method closely approximates a more rigorous solution but saves computational cost by two to three orders of magnitude. Further, a method of determining the optimal discrete state sequence in the switching state space model is invented that directly exploits the observation vector on a frame-by-frame basis and operates from left to right in time.

Patent Information

Applicants

Inventors

Specification

Documents