Audio coding using the best level wavelet packet transform and auditory masking

In this paper, we propose an audio coder that aims at high quality even in nonstationary segments of music signals. First, we divide input signals into subbands by using wavelet packet decomposition. We use a critical band approximate wavelet packet decomposition tree for efficient auditory masking. Dynamic wavelet packet decomposition is then used for effective signal representation. By choosing the basis from a wavelet packet tree, the best level method is used. The bases of the best level method has more entropy than that of the best basis method, but by choosing the appropriate frame length, it has similar entropy to that of the bases of the best basis method. Thus we can reduce side information. We propose an adaptive frame method whose criteria is the entropy of best level basis