Studying Performance For The Gain Filters Of 8kbit/S Ld-Acelp

Wu Shuhong,Zhang Gang
2011-01-01
Abstract:In order to meet ITU requirement that the delay is less than 5ms and coding rate is 8Kbit/s, we presented an 8Kbit/s 2.5ms low delay speech coding algorithm named LD-aCELP [1]. The algorithm used adaptive codebook and backward pitch detection. LD-aCELP depended on the Levinson-Durbin (L-D) algorithm to update gain filter coefficients. To develop the coding quality, we used three different methods which were the weighted L-S recursive filter, the finite memory recursive filter and the BP neural network to compare with L-D. The gain filter and its coefficients were obtained by a new scheme to estimate SNR. To evaluate performances of these gain filters using this method, the results are all better than LD-aCELP. The weighted L-S algorithm has the best effect, which is accordant with real speech coding. Its average segment SNR is higher than LD-aCELP about 0.720dB.
What problem does this paper attempt to address?