Research Of Active Interaction Technique For Human Intentions Under Gait Features

In recent years,more and more service equipment has been widely deployed in public places to provide corresponding services,reducing labor costs.The improvement of technique and the change of people’s life concept have put forward higher requirements for the intelligentization of service equipment.At present,the service equipment mainly adopts the traditional reactive interaction method,and can only provide the corresponding response after the user initiates an operation.In order to further attract users,improve the utilization rate of service equipment,and provide users with a more natural and humanized interactive experience,it is necessary to identify and understand the user’s intention in the "pre-interaction" stage based on the behavior status of pedestrians to initiate active interaction.This paper studies the active interaction technique of service equipment oriented to user intent under the gait features of pedestrians.The main researches are as follows:First,the human skeleton depth data of pedestrians in the "pre-interaction" stage were collected,and the behavior modes and interaction intentions of pedestrians in this stage were classified.By calculating and analyzing the depth data under different behavior modes,the gait features related to the intention were extracted,including the pedestrian’s position feature,velocity feature and direction feature based on the body turning angle,and the gait feature sequence was constructed.Secondly,the gait feature sequences were divided into different behavior frames according to the behavior modes of pedestrians,and the behavior frames were annotated with intentions,and the mapping relationship between the gait features and intentions was established.The validity of using the extracted gait features to identify intentions was verified by one-way ANOVA.Then use the labeled data to train the commonly used classifiers,and then selected the support vector machine to realize the position-velocity-direction-based intent recognition algorithm,and the recognition accuracy was 93.8%.Finally,an active interaction strategy was proposed to complete the design of the way and content of active interaction.Experiments was designed to compare the reactive interaction technique and the proposed intent-oriented active interaction technique,and evaluated the performance of the two technologies through quantitative analysis and qualitative analysis.The results show that the intention-oriented active interaction technique improves the utilization rate of service equipment and user conversion rate by 9.57% and30.10% respectively,and obtain better user evaluation.Based on the human skeleton depth data of pedestrians,this paper proposed an active interaction technique for the user’s intents under the gait features,and experiment was designed to evaluate the performance of the technique.The proposed technique achieved the experience goals of active interaction about the accuracy and non-perceptual,solved the problems of "if,when and how" in active interaction design,and provided research methods and design ideas for the interaction design of intelligent service equipment in the "pre-interaction" stage.