A Novel Sensor-Based Human Activity Recognition Method Based on Hybrid Feature Selection and Combinational Optimization

In recent years, sensor-based human activity recognition (HAR) has become a hot topic due to the advancement of sensing technologies, wireless communication technologies and nano-technologies. Since sensor signals are usually non-stationary quite noisy, both selecting discriminant feature representations finding out optimal parameters for algorithm play an important role enhanced performance robustness HAR system. However, most previous research focused on one them ignoring their interactions. Very few studies these two aspects simultaneously. Considering factors separately may lead inferior performance. This paper presents novel framework which can optimize set synchronously robust system A new hybrid selection methodology using game-theory based (GTFS) binary firefly (BFA), called GTFS-BFA, is proposed. GTFS-BFA combining evidence from filter wrapper methods. It consists phases, namely pre-selection phase re-selection phase. Pre-selection relies game-theory-based method, while uses (BFA) as method. The popular efficient kernel extreme learning machine (KELM) utilized classifier. experimental results indicate that proposed method obtain better comprehensive in terms four measures through comparison other existing methods daily dataset five body positions.