| Matching pursuit is a new algorithm for sparse decomposition of the signal,its sampling frequency is much lower than the Nyquist sampling frequency.The main purpose is sparse decomposition on a complete dictionary.The algorithm has been widely used in various fields.At present,the technology has gradually become a hot direction of audio and video coding technology.The content of this paper is the matching pursuit speech coding algorithm at a low bit rate.The traditional matching pursuit coding algorithm simplifies multiple optimal approximations into a single optimal approximation problem by a highly redundant dictionary space.As an effective audio codec algorithm,Ravelli algorithm although with high matching accuracy,the computational complexity also in a high value.Aiming at the problem that the dictionary of Ravelli algorithm is too large to lead to the longer time of matching,this paper proposes a matching pursuit optimization algorithm based on sparse adaptive speech coding technology at low complexity.Based on the MDCT dictionary adopted by Ravelli algorithm,the algorithm increases the Gabor dictionary which has the high time domain energy,raising the degree of matching between the dictionary and the signal;performing a energy-based preprocessing to select the higher energy portion of the same sample as the dictionary dimension.The matching algorithm is performed on each signal segment at each iteration times,which reduces the time-consuming calculation of the routing algorithm in traversing the whole signal frame.Experiments show that compared with the Ravelli algorithm the MUSHRA score is reduced of 0.73 at the same bit rate,and under normal circumstances,the computational complexity of the matching process is reduced to about 97%of the Ravelli algorithm.In conclusion,this paper aims to study the speech coding algorithm at a low complexity,and the corresponding solution is put forward by analyzing the problem that the matching algorithm exists in the matching process. |
PDF Full Text Request