引言安防感知的演进与核心痛点 | Introduction: The Evolution and Core Challenges of Security Sensing传统的安防系统如摄像头和门窗传感器主要扮演“事后记录”或“边界报警”的角色。它们往往在事件发生后才发出警报缺乏对空间内人员状态、行为意图的主动感知与预判能力。在智能家居和智慧养老场景下用户需要的是更智能、更人性化、更尊重隐私的“主动关怀”式安防。例如如何在不侵犯隐私的前提下判断独居老人是否长时间未活动或意外跌倒如何让家电设备理解人的存在与意图实现无感化、节能化的智能交互这些需求催生了新一代智能安防感知方案。Traditional security systems, such as cameras and door/window sensors, primarily serve the roles of “post-event recording” or “perimeter alarm.” They often trigger alerts only after an incident has occurred, lacking the ability to proactively sense and anticipate the status and behavioral intentions of occupants within a space. In smart home and elderly care scenarios, users demand more intelligent, humane, and privacy-respecting “active care” security. For instance, how can one determine if an elderly person living alone has been inactive for an extended period or has had a fall without infringing on privacy? How can home appliances understand human presence and intent to achieve seamless and energy-efficient intelligent interaction? These demands have given rise to a new generation of smart security sensing solutions.技术原理多维感知技术的融合与选择 | Technical Principles: Integration and Selection of Multi-Dimensional Sensing Technologies实现主动、非接触式的安防感知核心在于选择合适的传感技术。目前主流方案包括被动红外PIR、毫米波雷达和声波感知它们各有优劣适用于不同场景。被动红外PIR通过检测人体发出的红外热辐射变化来感知运动。其优点是成本低、功耗低但缺点显著无法检测静止或微动的人体易受环境温度干扰且无法判断距离、方向和人数。它仅适用于基础的“运动检测”。毫米波雷达通过发射并接收反射的毫米波信号可以精确检测物体的距离、速度、角度甚至微动如呼吸。其穿透性强不受光线、温度影响能实现真正的“存在感知”和“生命体征检测”。但传统雷达方案成本较高算法复杂且存在隐私担忧尽管不成像。声波感知以HomeSense™为代表这是一种创新性方案利用设备已有的扬声器和麦克风发射和接收超声波或特定频段声波。通过分析声波在空间内反射的多普勒效应和信道状态信息可以检测人体微动、接近、远离、跌倒等多种状态。其最大优势是零额外硬件成本直接复用设备的音频系统极大降低了方案集成门槛和BOM成本。同时它完全保护视觉隐私。对比总结对于需要高精度、全天候存在感知和生命体征监测如跌倒检测、离床判断的场景毫米波雷达是优选。而对于成本敏感、需快速集成到现有音频设备如智能音箱、电视、空调中实现近场感知、风随人动、防误唤醒等功能的场景声波感知方案HomeSense™展现出极高的性价比和实用性。Achieving proactive, contactless security sensing hinges on selecting the appropriate sensing technology. Current mainstream solutions include Passive Infrared (PIR), Millimeter Wave Radar, and Acoustic Sensing, each with its own strengths and weaknesses, suited for different scenarios.Passive Infrared (PIR): Detects motion by sensing changes in infrared heat radiation emitted by the human body. Its advantages are low cost and low power consumption, but it has significant drawbacks: it cannot detect stationary or slightly moving individuals, is susceptible to ambient temperature interference, and cannot determine distance, direction, or the number of people. It is only suitable for basic “motion detection.”Millimeter Wave Radar: By transmitting and receiving reflected millimeter-wave signals, it can precisely detect an object’s distance, speed, angle, and even micro-movements (such as breathing). It has strong penetration, is unaffected by light or temperature, and enables true “presence sensing” and “vital sign detection.” However, traditional radar solutions are relatively high-cost, involve complex algorithms, and raise privacy concerns (despite not capturing images).Acoustic Sensing (represented by HomeSense™): This is an innovative solution that utilizes a device’s existing speaker and microphone to emit and receive ultrasonic waves (or sound waves in specific frequency bands). By analyzing the Doppler effect and channel state information of sound wave reflections within a space, it can detect various human states such as micro-movement, approach, departure, and falls. Its greatest advantage is zero additional hardware cost, directly repurposing the device’s audio system, which significantly lowers the integration barrier and BOM cost. Simultaneously, it fully protects visual privacy.Comparative Summary: For scenarios requiring high-precision, all-weather presence sensing and vital sign monitoring (e.g., fall detection, bed exit monitoring), millimeter wave radar is the preferred choice. For cost-sensitive applications that require rapid integration into existing audio devices (e.g., smart speakers, TVs, air conditioners) to achieve near-field sensing, wind-following, anti-false-wakeup, and similar functions, acoustic sensing solutions (HomeSense™) demonstrate extremely high cost-effectiveness and practicality.核心产品与方案嵌入式AI感知引擎 | Core Products and Solutions: Embedded AI Perception Engines将先进的感知技术转化为可大规模商用的产品需要高度集成化、低功耗的硬件模组与高效算法。以行业领先的解决方案为例C2015 智能语音与感知AI芯片/算法IP这是一款高度集成的解决方案。它内置了高性能的语音识别ASRNPU和专门用于感知计算的NPU。其感知算法IP尤为出色能够在仅约 20KB的RAM 资源下稳定运行人体存在检测、轨迹跟踪、跌倒检测等复杂算法。这种极致的资源优化使其可以轻松部署在启英泰伦、晶晨Amlogic、全志Allwinner等主流语音AI芯片平台上无需更换主控仅通过软件升级或增加一颗低成本协处理器双芯片架构即可为设备赋予强大的“感知”能力通过标准的 UART或I2C接口 输出结构化感知结果。C0928 离线语音与存在感知二合一模组这是一款即插即用的标准化模组。它集成了高性能离线语音识别引擎和基于声波的存在感知引擎HomeSense™。单个模组即可实现“语音控制”“人在感知”两大核心功能。其典型感知能力可达检测距离约5米检测角度150°足以覆盖一个标准房间的大部分区域。家电厂商无需自行开发复杂的传感融合算法直接采用该模组可大幅缩短产品研发周期快速推出能“听懂”且“看懂”感知用户的智能产品。Transforming advanced sensing technologies into products suitable for large-scale commercialization requires highly integrated, low-power hardware modules and efficient algorithms. Taking industry-leading solutions as examples:C2015 Intelligent Voice and Perception AI Chip/Algorithm IP: This is a highly integrated solution. It incorporates a high-performance voice recognition (ASR) NPU and a dedicated NPU for perception computing. Its perception algorithm IP is particularly outstanding, capable of stably running complex algorithms for human presence detection, trajectory tracking, fall detection, etc., using only about 20KB of RAM resources. This extreme resource optimization allows it to be easily deployed on mainstream voice AI chip platforms such as those from Chipintelli, Amlogic, and Allwinner. Without changing the main controller, powerful “perception” capabilities can be added to devices merely through a software upgrade or by adding a low-cost coprocessor (dual-chip architecture), outputting structured perception results via standard UART or I2C interfaces.C0928 Offline Voice and Presence Sensing Combo Module: This is a plug-and-play standardized module. It integrates a high-performance offline voice recognition engine and an acoustic-based presence sensing engine (HomeSense™). A single module can achieve two core functions: “voice control” and “human presence sensing.” Its typical sensing capability reaches a detection range of approximately 5 meters with a 150° detection angle, sufficient to cover most areas of a standard room. Appliance manufacturers do not need to develop complex sensor fusion algorithms themselves. By directly adopting this module, they can significantly shorten product development cycles and quickly launch intelligent products that can “hear” and “see” (sense) users.应用案例赋能智能家居与智慧养老 | Application Cases: Empowering Smart Homes and Elderly Care基于上述方案智能安防感知正在众多场景中落地生根提升用户体验与安全。智能空调如美的、格力搭载C2015算法IP的空调可通过毫米波雷达或声波感知实时定位人体位置。实现“风避人吹”避免冷风直吹和“风随人动”送风跟随人员移动在提升舒适度的同时实现节能。夜间模式还可监测睡眠期间的微动调节风速温度。智能音箱如小米小爱、天猫精灵集成C0928模组的智能音箱可利用HomeSense™技术实现“近场唤醒”。只有当用户走近音箱