Abstract: A new code compression scheme was proposed to solve the power consumption problem in embedded systems.After analyzing the characteristics of instructions in target program, instructions will be combined and split, then the canonical Huffman algorithm is used to encode the modified instructions to generate look-up tables.Finally, code will be compressed and decompressed on the base of the corresponding relations between instructions and code words in look-up tables. According to the compression rate and power reduction rate of partial programs of embedded benchmarks compressed by simplescalar simulator, statistical data showed that the proposed algorithm can effectively save the storage space and reduce the power consumption of system.