Concept only · working name · not for sale
静かに、そばに。 守り · まもり
Mamori is a concept for a small tabletop device for an older relative’s home. It is designed to offer a help button, read the room air, speak to the person at home, work over a SIM even without home Wi-Fi, and keep its microphone off unless you can see that it is on. There is no camera.
The film’s ten captions, in order. Everything the film shows is design intent for a concept, not a tested product.
The story
Picture a mother in her eighties who lives alone in a quiet flat in Japan. Her children live in another city, or another country. They call on Sundays, and they wonder how she is in between.
Most ways to keep an eye on someone feel like being watched: cameras, constant check-ins, gadgets that treat a person like a patient. Mamori is designed to do the opposite. It would let a family stay close, and let her stay in charge of her own home.
In the room it would be a calm object on a shelf: soft fabric, light oak and one word on a small disc. It would read the room air, speak when asked, and offer one button for help. There is no camera, and the mic would be off unless its light is on.
The idea
Each one is a design intent, not a tested result. The limits sit right next to the promise.
1 Emergency warning
A press would sound an alarm in the room, turn the light ring red and show “Help called” on the status disc. If a family node is paired, it would send an alert there too, over Wi-Fi or the SIM.
Limit: only helps if the button can be reached. It is not an emergency service and there is no guaranteed response.
2 Living signals
PM2.5, CO2 and VOC, temperature and humidity, shown as simple notes like “air looks stale”.
Limit: room air readings, not health measurements.
3 Possible-fall alert (target)
Mamori would sense motion in the room through Wi-Fi channel-state sensing, with no camera. It is designed to flag a possible fall: a target to be validated in the pilot; not a medical device. A flag would go as an optional alert to the paired node. A partner’s separate sensor becomes optional.
Limit: not validated. It may miss real events or raise false alarms, and it depends on the room and the setup.
4 Optional call
Designed to call another paired node over home Wi-Fi or the SIM. The light ring would pulse amber and the status disc would say “Mic on”.
Limit: designed to be off while the shutter is closed. Not an emergency call line.
5 Full privacy
No camera and no image sensor in the concept. The mic LED is designed to sit on the mic’s power line, so no light would mean no mic. A small shutter on the side would cover the mic and cut its power. With the shutter closed it cannot listen, but it can still speak alerts. The wake word would run on the device; speech would go to the cloud only after the wake word, with the shutter open.
Limit: the circuit would need an independent check before anyone relies on it. Cloud speech means spoken requests leave the home.
Sensor map
About 90 mm across and about 89 mm tall, on a 24 mm oak base.
APoint 1 · Emergency warning
A small round help button on top. A press would sound a local alarm, turn the light ring red and show “Help called” on the status disc. If a family node is paired, it would also send an alert there.
Honest limit: only works if the person can reach and press it. It is not an emergency service; there is no guaranteed response.
Privacy, by hardware
Try it. The ring, the mic LED and the shutter tell the same story. This is a live demo of the design intent; nothing here uses your microphone.
Private. The microphone would have no power. It can still speak alerts.
The concept has no camera or image sensor, so it is designed to make no images. Sound would not be stored. The built-in Wi-Fi sensing reads changes in the Wi-Fi signal, not images. A partner’s sensor is optional, a separate device with its own privacy terms, which this concept does not control.
How it would work
A direct link between paired nodes is the intent, over home Wi-Fi or the SIM, so it is designed to work without home Wi-Fi. If nothing is paired, everything stays in the room.
Partner’s sensor (optional)separate, not required
Mamori, at homespeaks · Wi-Fi or SIM · mic off
Paired family nodea phone app or a second Mamori
Living signals
A mock of what the paired node might show. Plain words first, numbers second.
Living room · mock screen
Example values · illustrative only · not live data
Air looks a little stale. Opening a window could help.
PM2.5
12 µg/m³
Looks clear
CO2
840 ppm
A little stale
VOC
Low
Nothing to note
Temperature
22.5 °C
Comfortable
Humidity
48 %
Comfortable
These are room air readings, not health measurements. Low-cost sensors drift and need calibration; a spot near a heater skews them.
Specs
Design intent, not final engineering
| Body | Soft off-white fabric cube, about 90 mm across and about 89 mm tall; the fabric is also the speaker grille |
|---|---|
| Base | Light oak, 24 mm tall, with the nano-SIM tray underneath |
| Status disc | Round display of about 30 mm in a 40 mm clear disc, with an RGB ring around it; shows one word: Private, Mic on or Help called |
| Help button | 24 mm white button on top, with a thin persimmon edge |
| Mic shutter | On the right side; closed, it covers the mic and cuts its power |
| Sound | Speaker behind the fabric and a voice assistant: wake word on the device, then cloud speech |
| Wi-Fi sensing | Built in, using the device’s own Wi-Fi radio; designed to flag a possible fall: a target to be validated in the pilot; not a medical device |
| Mobile link | LTE-M or LTE Cat-1 with a nano-SIM, so it is designed to work without home Wi-Fi |
| Air sensors | PM2.5, CO2, VOC, temperature and humidity: room air, not health data |
| Camera | None. No camera and no image sensor |
Timeline
Indicative, not a commitment
Rough months from the start of prototyping. A plan, to be agreed, not a commitment. No dates are set. Each stage starts when the one before it is done.
Now
Design, look, this kit: functions, privacy rules, candidate parts, open questions.
Months 1 to 4
Works-like, looks-like, then EVT. Development boards and 3D-printed shells first; then the first custom board in the real housing (EVT): antennas tuned in the fabric, first Wi-Fi sensing, wake-word and cellular trials, first user feedback.
Months 4 to 7
Production-intent tools and fabric, reliability tests; radio (Giteki), terminal conformity for the cellular part, EMC (VCCI) and the adapter’s PSE; network operator device approval if the operator requires it.
Months 7 to 9
Pilot units from the production line in a few homes in Japan, with their families; the possible-fall flag (a target, not a medical device) and the voice assistant are to be validated here.
Launch follows, from month 10.
Full kit
The spec, the candidate parts, the manufacturing plan, Japan certification, the timeline and the risks. Concept, to be validated in the prototype. Nothing here is tested.
Honest answers
No. It is a concept, not a medical device, and it makes no medical claim. It would not replace a doctor, a carer or an emergency service.
Not as a proven feature. It is designed to flag a possible fall using Wi-Fi sensing: a target to be validated in the pilot; not a medical device. It may miss real events or raise false alarms, and how well it works would depend on the room and the setup. A partner’s separate sensor could be an optional extra input.
No. The concept has no camera and no image sensor, so it is designed to make no images to store or send.
The design intent is no: the mic LED would be wired on the mic’s power line, so the mic would not be able to work without the light on, and a small shutter on the side would cover the mic and cut its power. With the shutter closed it cannot listen, but it can still speak alerts. This hardware rule would need an independent check of the circuit before anyone relies on it. Sound would not be stored.
Not guaranteed. A press would sound a local alarm, which helps only if someone can hear it, and would send an alert to a paired family node if one is set up and the power and the Wi-Fi or mobile network link work. It is not an emergency service.
It would read the room air: PM2.5, CO2/VOC, temperature and humidity. Those are not health measurements.
No. This is a concept for discussion. There is nothing for sale, no account and no paid service. Mamori is a working name and has not been trademark-checked.