DesPatNet25: Data encryption standard cipher model for accurate automated construction site monitoring with sound signals. (1st May 2022)

• Record Type: Journal Article
• Authors: Akbal, Erhan; Barua, Prabal Datta; Dogan, Sengul; Tuncer, Turker; Acharya, U. Rajendra
• Is Part Of: Expert systems with applications. Volume 193(2022)
• Publication Date: 2022-05-01
• Subjects: DesPatNet25 -- Construction site monitoring -- Huge sound dataset -- ESC -- Vehicle identification using sound; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2021.116447
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: A multiple kernel based DES pattern is proposed. A new sound dataset was collected to detect CSM tasks. New hand-modeled learning model is proposed and it is named DesPatNet25. DesPatNet25 applied to the collected dataset and it reached 97.05% accuracy. Abstract: Nowadays, various construction site monitoring (CSM) models have been presented using sound signals. Many researchers have used deep learning (DL) networks to develop an accurate automated CSM model. These DL-based models require huge dataset to train the model and also such networks are complex. Hence, in this work, a novel hand-modeled automated system is developed using a public CSM sound dataset. The proposed model uses the first S-Box of the data encryption standard (DES) cipher as a feature generator by using two binary kernels. Using tent average pooling, sub-bands (compressed) sound signals are generated and the presented multiple kernelled DES pattern generates features from each signal. The proposed hand-modeled automated system extracts 25 feature vectors, hence it is named as DesPatNet25. The developed DesPatNet25 consists of: (i) feature vectors creation, (ii) feature selection using iterative neighborhood component analysis (INCA), and (iii) classification. Our proposed model attained accuracies of 96.77% and 97.05% using k-nearest neighbor (kNN) classifier with 10-fold cross-validation and hold-out validation (80:20 split ratio) techniques, respectively. These high classification accuracies clearly demonstrate the success of the DesPatNet25 model with sound signal classification for automated CSM tasks.