AI / LLMChina
← Sticker Wall
AGENT · DETAIL
Bathroom Fall Early Warning Radar Kit Monthly Subscription: Millimeter-wave non-inductive identification, 3 nursing homes generate 36,000 RMB monthly
Workflow: The input is point cloud frames continuously output by the ceiling-mounted 60GHz millimeter-wave radar in the bathroom,
AGENT
Key Fields
FIELD STAMPSIndustryAI / LLM
RegionChina（中国大陆）
ScaleSME
ChannelOnline
🔧 Workflow
The input is point cloud frames continuously output by the ceiling-mounted 60GHz millimeter-wave radar in the bathroom, referring to the 7×24-hour specifications of Aiqiangua UE care; processing is done by an edge gateway running a self-trained fall model to determine the posture change rate, and then calling ERNIE Bot to generate alert semantics, with similar solutions achieving an [email protected] of 89.3% on the RK3588 NPU; the output is instant alerts, nursing suggestions, and weekly report items; alerts are pushed in real-time, and weekly reports are summarized weekly; the closed-loop is completed after the caregiver arrives on-site to review and confirm, with false-positive samples fed back for retraining (specifications and case calibers).
🛠 Setup Requirements
It is necessary to purchase a 60GHz millimeter-wave radar development kit and a Raspberry Pi 4B or RK3588 edge gateway, deploy the self-trained fall detection model using an edge AI framework, and configure the WeChat Work bot pushing channel. Embedded Linux, Python, and basic model deployment capabilities are required, taking about 3 to 4 weeks to complete the model room and run through the entire chain.
🧰 Toolchain
- 🔧 60GHz millimeter-wave radar module (such as TI IWR6843 or Infineon BGT60)
- 🔧 Raspberry Pi 4B or RK3588 edge gateway
- 🔧 ERNIE Bot API (used to generate alert semantics and nursing suggestions)
- 🔧 WeChat Work bot or DingTalk bot pushing channel
💰 Revenue
① Senior care institution monthly subscription (main revenue): Nursing homes and daycare centers pay 8,000 to 12,000 RMB/institution per month. Signing 3 institutions = monthly revenue of 24,000 to 36,000 RMB. The weight of this line in total revenue has not been disclosed (numbers in the case, independently unverified; as of 2026); ② Hardware kit sales: Institutions purchase radar plus edge gateway per set, with a single set cost of 2,500 to 3,500 RMB, and the selling price has not been publicly stated (cost reported by the merchant, independently unverified), and its proportion is also vacant; ③ Additional institution sign-ups for revenue expansion: Each newly added institution only increases hardware amortization and about 300 RMB/month in communication and O&M costs, with a monthly subscription price of 8,000 to 12,000 RMB. How many more institutions were actually signed is unknown, and share data is unavailable; ④ Opportunity item - Home C-end subscription: Selling the same set of equipment to families on a monthly basis. Referring to homologous actual measurements, after edge computing optimization, the missed fall detection rate dropped from 35.2% to 7.8% (actual measurement estimation reported by media, independently unverified), and the exact share of this segment also has no figures.
💸 Cost
The hardware cost per set is about 2,500 to 3,500 RMB (radar module plus edge gateway), the large model API and communication cost per institution per month is about 300 RMB, and the total preliminary investment for the model room is about 6,000 RMB.
⏱ Time Investment
About 12 hours per week, including equipment status inspection, alert review, weekly report compilation, and renewal communication. Maintenance time gradually decreases after the system stabilizes.
🚀 Getting Started
The first step is to purchase a 60GHz millimeter-wave radar development board, use the manufacturer's SDK to read point cloud data and run the official fall detection demo, and then record simulated fall data of yourself or family members to fine-tune the model. The second step is to find a small local nursing home or day care center to run a free pilot, accumulate real-scene data, and then convert to a monthly subscription fee.
🔑 Keys to Success
- ✅ High-moisture environments such as bathrooms are not affected by visual occlusion, and radar recognition is stable with significantly reduced false positives
- ✅ Fully offline edge inference guarantees privacy, images do not leave the institution, and acceptance by family members and caregivers is high
- ✅ The monthly SaaS model turns one-time hardware sales into recurring cash flow, resulting in stronger institutional stickiness
- ✅ Select hardware components with an IP65 protection rating or above to ensure long-term stable operation in humid bathroom environments
⚠️ 风险
- ⚠️ Millimeter-wave radar has relatively weak recognition capabilities for static falls and backward falls; continuous collection of real fall postures is required to tune the model, otherwise missed alarms will lead to a loss of customer trust
- ⚠️ Senior care institutions are sensitive to new monthly subscription expenses. Once lower-priced competing products or local system integrators engage in price wars, the renewal rate may decline
- ⚠️ Changes in caregiver workflows require training coordination; initial unfamiliarity with operations may lead to delayed alert responses, affecting customer satisfaction
📌 Real Cases
- 📌 Developer TheDan open-sourced the 'ERNIE Guardian' system in the Volcano Engine ADG community, using 60GHz radar plus ERNIE Bot to achieve non-inductive elderly care, proving that the radar plus large model fall detection route can be implemented by individual developers
- 📌 Aiqiangua launched the UE care millimeter-wave radar fall alarm, mass-produced and sold for home and institutional scenarios, verifying the real demand for millimeter-wave fall alarm hardware in the elderly care market
- 📌 Moxun multi-modal algorithms have been implemented in senior care institutions. After edge computing optimization, the missed fall detection rate was reduced by 78%, verifying the commercial value of this technology in senior care scenarios