Skip to main content

Advanced signal processing and pattern identification

For multisensing to work, signals need to be synchronised. This is an orchestra.
These are not independent sensors.

We collect around 50 Mbytes of data per day in an average installation.
Our A.I. pinpoints life events in this multi-dimensional space, like for instance a shower.

Edge A.I.

Carefully designed A.I. always ends up into a matrix multiplication.
Complicated A.I. ends up on the Amazon cloud.

We are masters at designing small and optimised mathematical computing.
Even complex neural networks can be expressed as simple matrix multiplications.
This is our secret for keeping A.I. on the edge and not in the cloud.

Wavelet transforms experts

This is not a flower field but our A.I. at work.
Good mathematics is often beautiful.

We have developed proprietary wavelet analysis in real time for identification of sound and signal patterns.
With a low footprint in memory and processing power, we are able to characterize tiny details within a global context.

Ambient 16

Many sensors in a single 5x5 cm box, flat sensitive surface, any color. AI included.

temperature – humidity – air pressure – ndir co2 – light intensity and solar/artificial light detection – sound level ­and identification – vibration 
presence  – motion speed, direction ­­­and distance – draught air velocity ­– orientation – no hub – WiFi – portable, magnetic or as wall socket – Edge A.I.

Heatmap 32

Facial recognition is not a threat when it is not possible.

A low-resolution infrared radar for silhouettes and postures identification.
Activitity identification with neural network on chip. Fever detection.
Flow tracking for public areas and retail stores.

Nose 4

Because life does not always smell the way it should.

Complex organic molecules identification with electrochemical sensors.
5x5 cm. Embedded neural networking. Pee, poo, sewer, garbage, musty air, perspiration.