For removing noises from recorded ECG signal, adaptive notch filter (ANF) and independent component analysis (ICA) are used in this paper. In ANF, notch filter is obtained by adding bandpass filter (BPF) and voltage amplifier of unity gain. ANF is obtained by cascading of notch filter and adaptive filter which makes it more robust in respect of locating the poles and zeros on the respective constrained circle. On the other hand, ICA establishes the new coordinates which are non-orthogonal and statistically independent. It solves the problem of blind source separation (BSS). The novelty of this work is that for the first time ICA is used for pre-processing of variety of ECG signals with linear discriminant analysis (LDA) classifier/detector. The motivation behind to use LDA was that, it minimises the variance and maximises the class distance of the two variables by which chances of false detection becomes very low.