Chun-Cheng Peng

CHORD RECOGNITION USING NEURAL NETWORKS BASED ON PARTICLE SWARM OPTIMIZATION

A sequence of musical chords can facilitate musicians in music arrangement and accompaniment. To implement an intelligent system for chord recognition, in this paper we propose a novel approach using Artificial Neural Networks (ANN) trained by the Particle Swarm Optimization (PSO) technique and Backpropagation (BP) learning algorithm. All the training and testing data are generated from Musical Instrument Digital Interface (MIDI) symbolic data. Furthermore, in order to improve the recognition efficiency, the cadence is also included as an additional feature. Cadence is the structural punctuation of a melodic phrase and it is considered as an important feature for chord recognition. Experimental results of our proposed approach show that the addition of this feature improves significantly the recognition rate, and also that the ANN-PSO method outperforms ANN-BP in chord recognition. In addition, since preliminary experimental recognition rates are generally not stable enough, we further choose the optimal ANNs to propose a two-phase ANN model to ensemble the recognition results.

AN EFFICIENT EVOLUTIONARY NEURAL FUZZY CONTROLLER FOR THE INVERTED PENDULUM SYSTEM

In this article we propose an evolutionary neural fuzzy controller for the planetary train–type inverted pendulum system (IPS) and verify its effectiveness. The novel hybrid particle swarm optimization (HPSO) learning algorithm of the proposed controller is based on approaches of the fuzzy entropy clustering (FEC), the modified PSO (MPSO), and recursive singular value decomposition (RSVD). The FEC is applied to generate base particles and the MPSO is proposed to effectively improve the performance of the traditional PSO. There are mainly two different characteristics between the MPSO and its original version; that is, the initial parameters of the MPSO are calculated by an effective local approximation method (ELAM), and the global optimum is chosen by the multi-elites strategy (MES). In addition, we use the RSVD to determine the optimal consequent parameters of fuzzy rules, in order to reduce requirements of the computational time and space. Experimental results show that the proposed approach outperforms the proportional–integral–derivative (PID), PSO, and MPSO in terms of better abilities of tracking and noise rejection for planetary train–type IPS.

System Identification Using Fuzzy Cerebellar Model Articulation Controllers

Being an artificial neural network inspired by the cerebellum, the cerebellar model articulation controller (CMAC) was firstly developed in (Albus, 1975a, 1975b). With the advantages such as fast learning speed, high convergence rate, good generalization capability, and easier hardware implementation (Lin & Lee, 2009; Peng & Lin, 2011), the CMAC has been successfully applied to many fields; for example, identification (Lee et al., 2004), image coding (Iiguni, 1996), ultrasonic motors (Leu et al., 2010), grey relational analysis (Chang et al., 2010), pattern recognition (Glanz et al., 1991), robot control (Harmon et al., 2005; Mese, 2003; Miller et al., 1990), signal processing (Kolcz & Allinson, 1994), and diagnosis (Hung & Wang, 2004; Wang & Jiang, 2004). However, there are three main drawbacks of Albus’ CMAC, i.e., larger required computing memory (Lee et al., 2007; Leu et al., 2010; Lin et al., 2008)), relatively poor ability of function approximation (Commuri & Lewis, 1997; Guo et al., 2002; Ker et al., 1997), and difficulty of adaptively selecting structural parameters (Hwang & Lin, 1998; Lee et al., 2003).

A self-organizing neural network using hierarchical particle swarm optimization

This paper introduces a hierarchical particle swarm optimization (HPSO) algorithm strategy for self-organizing neural network design. The proposed CHPSO can determine the structure of the neural network and tune the parameters in the neural network automatically. The structure learning is based on the genetic algorithm (GA) and the parameter learning is based on the particle swarm optimization (PSO). The advantages of the proposed learning algorithm can obtain fine structure and performance for neural network (NN). The prediction of simulation example has been given to illustrate the performance and effectiveness of the proposed model.

Chord recognition using neural networks based on Particle Swarm Optimization

A sequence of musical chords can facilitate musicians in music arrangement and accompaniment. To implement an intelligent system for chord recognition, in this paper we propose a novel approach using Artificial Neural Networks (ANN) trained by the Particle Swarm Optimization (PSO) technique and Backpropagation (BP) learning algorithm. All the training and testing data are generated from Musical Instrument Digital Interface (MIDI) symbolic data. Furthermore, in order to improve the recognition efficiency, the cadence is also included as an additional feature. Cadence is the structural punctuation of a melodic phrase and it is considered as an important feature for chord recognition. Experimental results of our proposed approach show that the addition of this feature improves significantly the recognition rate, and also that the ANN-PSO method outperforms ANN-BP in chord recognition. In addition, since preliminary experimental recognition rates are generally not stable enough, we further choose the optimal ANNs to propose a two-phase ANN model to ensemble the recognition results.

An efficient Symbiotic Taguchi-based Differential Evolution for neuro-fuzzy network design

In this paper, we proposed a functional-link-based neural fuzzy network to improve the traditional TSK-type neural fuzzy network. Besides, an efficient evolutionary learning algorithm, called the Symbiotic Taguchi-based Modified Differential Evolution (STMDE), is proposed for the neural fuzzy networks design. Firstly, in order to avoid trapping in a local optimal solution and to ensure the searching capability of near global optimal solution, the STMDE adopts the Taguchi method to effectively search towards the best individual and employs an adaptive parameter control to adjust scaling factor which is called the Taguchi method. Moreover, the proposed STMDE introduces the concept of symbiotic evolution to improve the individual structure. Unlike the traditional individual that uses each one in a population as a full solution to a given problem, symbiotic evolution assumes that each individual in a population represents only a partial solution, while complex solutions combine several individuals in the population.