Adapting horizon size in finite impulse response filtering through switching extensible FIR filter bank

In finite impulse response (FIR) filtering, horizon size (memory size) is an important parameter that influences estimation performance. In this paper, in order to improve the estimation performance of a linear FIR filter, we propose a new method to manage horizon size. The proposed method adapts (adjusts) horizon size based on the likelihood of observation and it is called switching extensible FIR filter bank (SEFFB). The SEFFB exhibits smaller estimation errors than a single FIR filter that uses the best constant horizon size.