研究方向 / Research
自然接口
Natural Interface
自然动作意图理解 | 副语言增强语音交互 | 情感交互 | 连续行为感知 | 移动、穿戴、车机、家居
Natural action intention understanding | Paralanguage enhanced speech interaction | Emotional interaction | Continuous behavior perception | Car machine, Home, Mobile, Wearable
针对自然人机交互的需求和挑战,研究多模态自然交互接口引擎:为不同场景下个性化的交互应用提供语音、触控、手势、姿态、视线、生物特征等多种交互通道上的自然交互输入能力和视听触呈现等自然交互输出能力,通过结合对用户先验知识的分析和建模,以及多模态自然表达的数据处理,在自然交互接口上实现准确的交互意图推理,扩充了意图推理方法的可扩展性和可解释性,实现从人适应机向机适应人的转变,显著提升交互性能。
In view of the needs and challenges of natural human-computer interaction, we study the natural interaction input capability on various interaction channels such as voice, touch, gesture, posture, line of sight and biological features, the natural interaction output capability such as audio-visual touch presentation, and process multimodal natural expression data, realizing accurate interaction intention reasoning on the natural interaction interface, the scalability and interpretability of intention reasoning method, the transformation from human adaptation to machine adaptation to human, and significantly improving the interaction performance.
高效交互
Efficient Interaction
交互任务图谱 | 情境感知 | 信息无障碍 | 界面自适应
Interactive task map | Context awareness | Information accessibility | Interface adaptation
面向万物互联时代下更广泛的交互场景,研究高效交互的计算原理和方法,提供适应性、学习性和主动性的交互性能,将人机交互从"控制"推向"理解"。根据用户状态和情境信息生成交互界面,解决交互决策心理模型的计算表示、自然运动控制能力的个性化建模方法和智能交互界面的构建方法三个科学问题。
Facing a wider range of interaction scenarios in the era of Internet of things, we study the computing principles and methods of efficient interaction to provide adaptability, learning and initiative of interaction. According to the user state and situational information, the optimized interactive interface is generated, and three scientific problems are solved: the computational representation of the psychological model of interactive decision-making, the personalized modeling method of natural motion control ability, and the modeling and semantic representation of interactive tasks. An intelligent interactive interface management system is constructed.
人机协同
Human-machine Coordination
黑盒算法可解释生成 | 人机信任建立与度量 | 多模态信息的理解与生成
Black box algorithm interpretable generation | Human-machine trust establishment and measurement | Understanding and generation of multimodal information
致力于将人类智慧与人工智能这两类异质的智能形态有机融合以期获得更高水平的总体智能,研究提高人机交互效率的方法,即如何将人的智慧和意图更自然高效地传达给机器,以及机器的运算结果如何让人更容易理解;探索人机两类异质智能配合工作的规律,研究AI与人智能的本质差异,人和AI的配合、互学习和互启发机制等。
Human-machine coordination is committed to the organic integration of two heterogeneous intelligent forms, human intelligence and artificial intelligence, in order to obtain a higher level of overall intelligence, and study the methods to improve the efficiency of human-computer interaction, that is, how to convey human intelligence and intention to the machine more naturally and efficiently, and how to make the calculation results of the machine easier for people to understand. The law of human-computer heterogeneous intelligence cooperation is explored. The essential differences between AI and human intelligence, and the cooperation, mutual learning and mutual inspiration mechanism between human and AI are studied.