Worker activity recognition is a crucial area of research with significant applications in manufacturing, healthcare, and security. Traditional methods relying on computer vision-based algorithms often demand extensive infrastructure, such as video cameras. This study presents a more efficient approach by utilizing data from embedded sensors in smartphones and microcontrollers. These wearable devices capture worker activities, including walking, sitting, standing, laying, and climbing stairs, through accelerometer sensors. The data is transmitted and monitored via an IoT platform, where machine learning techniques are employed for activity recognition and classification. The k-NN and Naïve Bayes classifiers achieved the highest accuracy rates, with 77.8% for smartphones and 76.1% for microcontrollers. This system provides a practical solution for recognizing and monitoring worker activities, with potential applications in factories for enhancing health, safety, and efficiency.