智能手表能够让用户方便的快速获取各种信息与执行各种快捷操作，这使得智 能手表的价值被越来越多的用户认可。然而目前的智能手表的交互能力依然有限， 提升智能手表的应用价值需要扩展其输入能力以支持可能需要的各种交互。本文 提出 SonarWatch 技术，这是一种基于声场传感的新颖传感与交互技术。SonarWatch 将智能手表的交互空间从手表屏幕上扩展到手表周围的区域，从而大大扩展了其感 知与输入空间。该技术的主要传感能力基于手表本身传感器所构建的声场(Sound Field)实现，在手表内建的惯性测量单元(Inertial Measurement Unit, IMU)所提供的 运动信号的辅助下，对 12 种交互手势的识别准确率达到了 93.7%，对 5 种单手交互 手势的识别准确率达到了 97.6%。通过系统的评估，本文测试并分析了 SonarWatch 的传感范围，感知灵敏度和感知局限性，充分的探索了手表声场感知的边界与交 互能力。SonarWatch 技术为智能手表的交互提供了各种各样的可能性。该技术支 持用户主动的单手/双手交互、触摸/非触摸交互以及用户日常的隐式状态识别，基 本覆盖了智能手表交互相关的研究中所提出的各种交互模态。由于 SonarWatch 仅 依赖于智能手表中已经存在的传感器，从而使其具有很高的实用价值和快速部署 的能力。本文的研究设计与开发了 SonarWatch 技术的硬件系统与算法模型，并在 不同应用场景与部署条件下评估了 SonarWatch 的整体性能和能耗。

Smartwatches allow users to conveniently and quickly obtain various information and perform multiple quick operations, which makes the value of smartwatches recog- nized by more and more users. However, the interaction capabilities of current smart- watches are still limited, and enhancing the application value of smartwatches requires expanding input capabilities to support various required interactions. This paper proposes SonarWatch technology, a novel sensing and interaction technology based on sound field sensing. SonarWatch expands the interaction space of the smartwatch from the screen to the area around the watch, greatly expanding its perception and input space. The primary sensing capability is realized based on the sound field built by the watch’s sensor. With the assistance of the motion signal provided by the built-in inertial measurement unit (IMU), 12 kinds of interactive gestures can be detected. The recognition accuracy rate reached 93.7%. The recognition accuracy rate for five same-side hand gestures reached 97.6%. Through systematic evaluation, this paper tests and analyzes SonarWatch’s sensing range, perceptual sensitivity, and perceptual limitations, then fully explores the boundaries and interaction capabilities of the watch’s sound field sensing. SonarWatch technology offers a wide variety of possibilities for smartwatch interaction. This technology supports the user’s explicit one-handed/two-handed interaction, touch/non-touch interaction, and daily implicit state recognition, covering various interaction modes proposed in past research related to smartwatch interaction. Since SonarWatch only relies on the sensors already present in the smartwatch, it is e?icient and capable of rapid deployment. The research of this paper designs and develops the hardware system and algorithm model of SonarWatch and evaluates the overall performance and energy consumption of SonarWatch under different application scenarios and deployment conditions.