Research On The Prediction Model Of Chaotic Speech Signal Based On Artificial Bee Colony Algorithm

Most of the engineering applications and optimization problems in real life are somewhat non-linear and difficult to determine.Traditional mathematics-based optimization methods are difficult to achieve an ideal result when dealing with these issues.Some researchers observed the habits of group organisms in the natural world,and discovered the intelligent behaviors of information exchange and cooperation among individuals in group activities.Based on this idea,the group intelligent optimization algorithm that uses mathematical process to imitate communication and cooperation among populations has become an important method to solve the optimization problems in many fields.When the chaotic speech signal studied in this paper is used to build a predictive model,the model is more complex and difficult to optimize by mathematical methods.Therefore,the group intelligent algorithm provides a new idea for our research.The traditional linear analysis and prediction methods can obtain better forecasting results,but due to the nature of speech nonlinearity,the prediction has certain limitations.The key to nonlinear speech prediction is to capture the chaotic features of speech,and based on this feature,a predictive model with good prediction effect is obtained.Therefore,in order to ensure the effect,the algorithm must be used to optimize the model.Chaotic speech signals have a high degree of complexity and nonlinearity,and it is difficult to achieve a considerable optimization effect through conventional mathematical methods.Artificial Bee Colony Algorithm(ABC)is an optimization algorithm for individual intelligence collaboration,information exchange,and foraging by bee colonies.A single individual has an impact on the entire population during the iterative process,and there is a good balance between exploration and development throughout the search process.Building a predictive model is a key step in speech signal processing.This paper analyzes and predicts speech signals based on chaos theory.In the phase space,the speech sequence is mapped as a continuous phase point and a connection is established between the phase points.And according to the chaotic characteristics of speech,it analyzes the speech to solve the random,nonlinear and unpredictable problems of the speech signal.In this paper,based on systematically analyzing the chaotic characteristics of speech,a second-order truncated Volterra model is used to construct nonlinear speech processes.The model was determined and optimized by the adaptive artificial bee colony algorithm based on global optimal guidance.And verify the effect of the predictive model on single-frame signals,words,sentence signals,and Lorentz sequences.The main work of this paper is as follows:1.Voice chaos test and analysis.The Lyapunov exponent of the preprocessed speech signal is used to determine the chaotic characteristics of the signal,and the delay time,the embedding dimension,and the continuous speech sequence are converted into phase points under phase space.2.Improved artificial bee colony algorithm.This paper proposes a global optimal adaptive artificial bee colony algorithm for the problems of slow convergence and solution accuracy of bee colony algorithm.In the course of each iterative calculation,the current best position is obtained to guide the next bee expansion search,and a random position is introduced in the search process to ensure the diversity of solutions;and a random number is generated using a chaotic map when the bee colony algorithm is initialized.After the improvement,the convergence speed of the algorithm is greatly improved,and the solution accuracy is good under the high dimensional and multi-peak test functions.3.Construct a chaotic speech signal prediction model.Based on the current chaotic characteristics of speech,based on the second-order truncated Volterra and AGABC algorithm,the structure and parameters of the prediction model are solved.The mean square error between the predicted value and the actual value is used as the fitness function for iterative calculation to obtain an explicit expression.A model that is easy to solve and has good prediction results.4.Experimental verification and analysis.The single-frame,word-sentence signal and Lorentz sequence prediction results were verified and compared,and the mean square error and waveform comparison were evaluated.And analyze the causes of the difference in the effect before and after,and draw conclusions.