Fast Huffman decoding algorithm by multiple-bit length search scheme for MPEG-2/4 AAC

Huffman coding is essential to the MPEG-2/4 Advanced Audio Coding (AAC) standard. The process of Huffman decoding requires a long delay because much computational time is consumed by an excessive number of searches for decoded symbols. In this paper, a fast Huffman decoding algorithm is proposed that reduces the number of searches required. The algorithm uses a search scheme in which a 5-bit length is used for each search basis to decode symbols. Experimental results show that a 5-bit search scheme can be hit with a probability of 70% to decode the symbols. Thus, the proposed method can reduce the number of search iterations, reducing the processing delay and increasing the processing efficiency for Huffman decoding. Compared with other algorithms for Huffman decoding, the proposed algorithm can reduce the number of search iterations by 35% to 73%, and the number of instructions by 32% to 60%.