Another kind of methods is data pre-processing like normalization and weighted values. Involving both feature selection and data pre-processing, a novel layered hidden Markov model (LHMM) was proposed to detect cardiac arrhythmia that is two layers of HMMs, the Layer ใ HMM extracts the features of ECG waveform and human activities, and the Layer 2 HMM classifies the type of cardiac arrhythmias. On the other hand, rule-based algorithms and the deterministic automaton relay on medical knowledge.