Analysis of human motion is studied in the context of 2D video image estimation. An adaptive human motion estimation filter with integrated phase compensator is designed to extract the relevant features from the human motion information through video samples. The use of phase compensator ensures that the algorithm performs well for both low and high frequency human motion estimation. An extended principal component analysis is applied to the two dimensional situation for dimensional reduction and results obtained, prove to be superior, compared to reported schemes. A correlation extractor is emulated in the hardware and extracts the interframe pixel correlation values and is spatial dependent. The implemented algorithm is scalable and generic and cause least modifications on the representation coefficients. Thus, it can tolerate increased variations on the motion appearance and keep the basis representation vector unaffected. Implementation results with real sequences are evaluated. The artefacts considered in the design include (i) Effects of expression, (ii) Illumination and (iii) Occlusion variations.