Discrete Wavelet Transform for Generative Adversarial Network to Identify Drivers Using Gyroscope and Accelerometer Sensors

Driver identification is a central research area in intelligent transportation systems, with applications commercial freight transport and usage-based insurance. One way to perform the use smartphones as main sensor devices. After extracting features from smartphone-embedded sensors, various machine learning methods can be used identify driver. However, accuracy often degrades number of drivers increases. This paper uses Generative Adversarial Network (GAN) for data augmentation obtain driver algorithm that maintains excellent performance also when Since GAN diversifies drivers’ data, it makes possible apply on larger without overfitting. Although GANs are commonly image processing augmentation, their driving signal novel. GAN’s training raw signals diverges. challenge solved by getting Discrete Wavelet Transform (DWT) before feeding GAN. Our experiments prove usefulness model generating emanating DWT smartphones’ accelerometer gyroscope signals. collecting augmented histograms along overlapped windows fed covered Stacked Generalization Method (SGM). The presented hybrid GAN-SGM approach identifies 97% accuracy, 98% precision, recall, F1-measure outperforms standard utilize extracted statistical, spectral, temporal approaches.