Novel neural network scheme composed of prediction and correlations
Yumi Takizawa, Atsushi Fukasawa · International Conference on Systems · 2009
This paper describes novel neural network scheme to minimize errors of prediction between observation and estimation. This scheme is discriminated from the conventional principle of learning methods. The proposed scheme is applied for the problem of time-space analysis of multiple sound source locations. The number of source are assumed 8 sound in a frame of 5 (s), and space size of sensors are 1(km) squared. The proposed scheme is proved with well coincidence of locations between correct and estimated positions.