In this chapter the continuous wavelet transform (CWT), based on a complex analysing function, is applied to characterize local symmetry of signals, and it is used for ECG arrhythmia analysis. The first part of this chapter is more theoretical. The behaviour of CWT square modulus of a regular signal f(t) when the scale parameter goes to zero is studied. For a signal with local symmetry properties, the phase behaviour of its CWT is also examined. These results are then extended, under some conditions, to signal without local symmetries. The second part is more experimental and numerical examples on simulated data illustrate the mathematical results. Finally, the use of these properties is considered in automatic ECG recognition and identification by means of hidden Markov models (HMMs). The presentation emphasizes how a suitable parameter vector, corresponding to the input observation sequence of the Markov chain, can be built and applied.