Research On Face Detection And Recognition Algorithm For Fitness Service Robots

With the continuous development of modern science and technology,artificial intelligence has been spreading to various fields of production and life,giving rise to one technological revolution after another.Service robots for the fitness fields are also gradually developing with the support of artificial intelligence technology.Human health issues have been widely concerned by society,and effective exercise is an important approach to ensure human health.Traditional fitness models often rely on the guidance of fitness coaches.In contrast,fitness service robots are increasingly popular among the general public due to their ability to develop personalized fitness plans for people.Face detection and recognition is an important technology for fitness service robots to quickly develop personalized solutions for users and simplify the interaction process.Fitness service robots not only need to store a large amount of personnel information,but also need to make quick and accurate responses.Therefore,designing face detection and recognition technology with good real-time performance and high recognition accuracy is of great engineering significance.This article conducts research on methods related to face detection and recognition based on fitness service robots.The main research work is as follows:(1)This chapter studied an improved face detection algorithm based on the YOLO(You Only Look Once)framework of object detection neural network.In response to the issue that the performance of the fitness service robot development board is inferior to that of the computer,and the real-time detection needs to be improved,a lightweight improvement has been made to the YOLO object detection algorithm.Firstly,the backbone network was replaced,greatly compressing the model size.Secondly,the accuracy of detection is ensured by introducing the Shuffle Attention mechanism.The experimental results showed that the improved YOLO face detection algorithm had better accuracy and real-time performance compared to the original algorithm.(2)This chapter studies the Face Net face recognition algorithm based on Shuffle Netv2.On the basis of the original Face Net model,the Inception Res Net V1 was modified to Shuffle Netv2,at the same time,the loss function was improved.Finally,the new algorithm was trained on the CASIA-Web Face dataset and tested on the LFW public dataset.The experimental results showed that the improved Face Net can effectively enhance the performance of facial recognition.(3)This chapter introduces the platform construction of fitness service robots,including hardware components,software architecture,etc.Firstly,the structure of the fitness service robot,the principle of high-precision height measurement,and the principle of heart rate and blood oxygen measurement were introduced.Secondly,the principle and implementation of voice interaction were introduced in detail.Finally,the paper conducted experiments on fitness service robots based on real scenario.The experimental results showed that fitness service robots could accurately detect and recognize users’ faces,and obtained the physiological data such as height and weight,engage in voice interaction,and provide accurate personalized suggestions.