A method based on interpretable machine learning for recognizing the intensity of human engagement intention
Authors:
Open Access
- 13 February 2023
- journal article
- research article
- Published by Springer Nature in Scientific Reports
- Vol. 13 (1) , 1-14
Abstract
To interact with humans more precisely and naturally, social robots need to “perceive” human engagement intention, especially need to recognize the main interaction person in multi-person interaction scenarios. By analyzing the intensity of human engagement intention (IHEI), social robots can distinguish the intention of different persons. Most existing research in this field mainly focus on analyzing whether a person has the intention to interact with the robot while lack of analysis of IHEI. In this regard, this paper proposes an approach for recognizing the engagement intention intensity. Four categories of visual features, including line of sight, head pose, distance and expression of human, are captured, and a CatBoost-based machine learning model is applied to train an optimal classifier for predicting the IHEI on the dataset. The experimental results show that this classifier can effectively predict the IHEI that can be applied into real human–robot interaction scenarios. Moreover, the proposed model is an interpretable machine learning model, where interpretability analysis on the trained classifier has been done to explore the deep associations between input features and engagement intention, thereby providing robust and effective robot social decision-making.Keywords
Funding Information
- Graduate Innovation Project of Chongqing University of Technology (clgycx20201010)
- Postdoctoral Science Foundation Program of Chongqing Science and Technology Bureau (No. CSTB2022NSCQ-BHX0674)
- Youth Project of Science and Technology Research Program of Chongqing Education Commission of China (No. KJQN202101131)
- Cooperative Project between universities in Chongqing and affiliated institutes of Chinese Academy of Sciences (No. HZ2021011)
This publication has 33 references indexed in Scilit:
- Starting engagement detection towards a companion robot using multimodal featuresRobotics and Autonomous Systems, 2016
- Designing social greetings in human robot interaction2014
- Toward understanding social cues and signals in human–robot interaction: effects of robot gaze and proxemic behaviorFrontiers in Psychology, 2013
- Distance-Based Computational Models for Facilitating Robot Interaction with ChildrenJournal of Human-Robot Interaction, 2012
- Pearson Correlation Coefficient2009
- Understanding human intentions via hidden markov models in autonomous mobile robots2008
- Recognizing action units for facial expression analysisIEEE Transactions on Pattern Analysis and Machine Intelligence, 2001
- Support-vector networksMachine Learning, 1995
- Multilayer feedforward networks are universal approximatorsNeural Networks, 1989
- Proxemics [and Comments and Replies]Current Anthropology, 1968