Kernel Estimation of Volterra Using an Adaptive Artificial Bee Colony Optimization and Its Application to Speech Signal Multi-Step Prediction

IEEE Access

Yumei Zhang¹  Chengling Yan²  Xiaojun Wu³  Yuying Rong⁴  Jiangshan Liu⁴ 

[1] Key Laboratory of Modern Teaching Technology, Ministry of Education, Shaanxi Normal University, Xi&x2019;School of Computer Science, Shaanxi Normal University, Xi&x2019;School of Environmental Sciences, Shandong University, Qingdao, China;an, China;

关键词: Speech signal;    kernel estimation;    Volterra model;    ABC algorithm;    multi-step prediction;   

DOI  :  10.1109/ACCESS.2018.2880280

来源: DOAJ

【 摘 要 】

In order to solve parameters selection problem when applying recursive least square (RLS), least mean square (LMS) or normalized LMS (NLMS) algorithms to estimate kernels of second-order Volterra filter (SOVF), a novel adaptive gbest-guide artificial bee colony (AGABC) optimization algorithm is used to derive kernels of Volterra, that is a type of the AGABC-SOVF prediction model with an explicit configuration for speech signal is proposed. The AGABC algorithm modifies the solution search equation of ABC algorithm and combines the best solution with neighborhood information at present iteration, which not only ensures the exploration of the global optimization algorithm but also improves the exploitation. The AGABC-SOVF model is performed to predict speech signal series of the given English phonemes, sentences, and chaotic time series. Simulation results based on benchmark function show that AGABC algorithm performs faster convergence in achieving higher quality solutions than original ABC and other improved ABC algorithms. Prediction results of applying the AGABC-SOVF model to multi-step predictions for Lorenz time series reveal its stability and convergence properties. For the measured multi-frame speech signals, prediction accuracy and length of multi-step prediction using the AGABC-SOVF model are better than that of the ABC-SOVF model. The AGABC-SOVF model can better predict chaotic time series and the real measured speech signal series.

【 授权许可】

Unknown