Pattern recognition approaches are commonly adopted by many wireless sensor network applications due to their traditional effectiveness in detecting and recognising events. Nonetheless, employing off-the-shelf pattern recognition approaches is generally infeasible. This paper presents a new energy efficient distributed pattern recognition approach based on information processing and sleep mode paradigms of sensor networks. Events of interest are recognised by the cooperative operations of the sensor nodes instead of depending on a base-station. Moreover, the aim of the approach is to attain effective event recognition performance whilst prolonging the network's longevity by switching off sensor nodes and processing sensory data within the network.