Development of an LSTM-Based Model with Attention Mechanism for Detection of Traffic’s Officer Hand Gestures for Autonomous Vehicles in Ecuador

Road traffic accidents are a significant global public health concern, with around 1.35 million deaths reported every year. According to the WHO (World Health Organization’s) in the Americas region, drivers and passengers of four-wheeled vehicles account for approximately 34% of road traffic deaths, while in Ecuador, around 1.1k people die from four-wheeled vehicle accidents, and around 4.2k fatalities are pedestrians. The potential for autonomous vehicles (AVs) to transfer critical safety tasks from humans to machines provides a solution to reducing road deaths. However, the realization of AV safety benefits requires technological advancements, public perception, and adoption rates. Deep learning and machine learning techniques have become fundamental tools in the research and development of AVs to detect various objects in their environment, such as vehicles, pedestrians, animals, and traffic gestures. However, the detection of traffic officers’ hand gestures by AVs has not been extensively explored. This article proposes the implementation of deep learning and LSTM (Long Short-Term Memory) techniques to estimate the gestures of a traffic officer that an autonomous vehicle must interpret. The experimental results demonstrate promising outcomes with an accuracy of 0.80. These findings showcase the effectiveness of the proposed model in achieving accurate predictions and highlight its potential for practical applications in the field.