We present a technique to find the biological function of protein from its primary sequence. Currently used protein classification methods make use of multiple alignments. We use signal-processing features obtained from the primary sequence of the protein, to predict its biological function. The primary sequence of protein is converted to signals based on the encoding of biochemical properties like hydrophobicity, solubility, molecular weight of constituent amino acids. Signal processing features like complexity, mobility and fractal dimension are extracted from the obtained signals. Studies are conducted for lipase, protease and isomerase of length between 100 and 200 amino acids.