Journal of Control Science and Engineering 1 (2015) 21-34 doi: 10.17265/2328-2231/2015.01.003 Comparison between Classical and Intelligent Identification Systems for Classification of Gait Events Carlos Galván-Duque 1 , Ricardo Zavala-Yoé 2 , Gerardo Rodríguez-Reyes 1 , Felipe Mendoza-Cruz 1 , MichelínÁlvarez-Camacho 1 and Ricardo Ramírez-Mendoza 2 1. Orthotics and Prosthetics Laboratory, Instituto Nacional de Rehabilitación, Mexico City, Mexico 2. School of Engineering and Architecture Graduate Studies, Tecnológico de Monterrey, Mexico City Campus, Mexico Abstract: Gait event detection is important for diagnosis and evaluation. This is a challenging endeavor due to subjectivity, high amount of data, among other problems. ANFIS (Artificial Neural Fuzzy Inference Systems), ARX (Autoregressive Models with Exogenous Variables), OE (Output Error models), NARX (Nonlinear Autoregressive Models with Exogenous Variables) and models based on NN (neural networks) were developed in order to detect gait events without the problems mentioned. The objective was to compare developed models’ performance and determinate the most suitable model for gait events detection. Knee joint angle, heel foot switch and toe foot switch during normal walking in a treadmill were collected from a healthy volunteer. Gait events were classified by three experts in human motion. Experts’ mean classification was obtained and all models were trained and tested with the collected data and experts’ mean classification. Fit percentage was obtained to evaluate models performance. Fit percentages were: ANFIS: 79.49%, ARX: 68.8%, OE: 71.39%, NARX: 88.59%, NNARX: 67.66%, NNRARX: 68.25% and NNARMAX: 54.71%. NARX had the best performance for gait events classification. For ARX and OE, previous filtering is needed. NN’s models showed the best performance for high frequency components. ANFIS and NARX were able to integrate criteria from three experts for gait analysis. NARX and ANFIS are suitable for gait event identification. Test with additional subjects is needed. Keywords: Gait analysis, biomechanics, system identification. 1. Introduction Since walking is a pattern of motion, diagnosis of the patient’s difficulties depends on an accurate description of the actions occurring at each joint. Traditionally, the method used for such description has been observed the patients gait. While performing the observation in a systematic manner results in more agreement among observers, there still is disagreement on details. An alternate approach is quantitated documentation of the person’s performance with reliable instrumentation that provides a permanent record of fact [1]. However, the analysis of quantitative data has been a challenging endeavour [2]. The high amount of data, nonlinear dependencies, inter-subject and intra-subject Corresponding author: Carlos Galván-Duque, B.S., research field: orthotics and prosthetics, E-mail: cgalvanduque@yahoo. com.mx. variability, among others, are typical problems when motion analysis is performed. These complexities are compounded by long recording times in gait laboratories, and increasing patient populations result in late diagnosis， leading to an increased risk of disorder progression and further complications [3]. CI (Computational Intelligence) is a fusion of learning mechanisms and computing, specifically suited for powerful decision systems capable of interpreting and processing large volumes of data [3]. ANFIS (Artificial Neural Fuzzy Inference Systems), ARX (Autoregressive Models with Exogenous Variables), NARX (Nonlinear Autoregressive Models with Exogenous Variables) and OE (Output Error models) are four of many techniques that can be used for pattern recognition, therefore, are suitable for gait analysis. With CI, it is possible to model a biomechanical system by learning data relationships between inputs and outputs possibly corrupted by D DAVID PUBLISHING