Skip to main content

Detecting activities from data

50 environmental values every 10 seconds.

The data showcased here were collected in a bathroom over a period of 32 hours.

Our artificial intelligence analyzes these time series, producing structured descriptions of household activities, which in turn enrich our statistics. These insights power our dashboards and enable our chatbot to provide realistic commentary on the activities.

We share this behind-the-scenes data to give you a glimpse into our process. However, as our valued customer, you won't need to see or analyze this data yourself.
 

Motion

Pyroelectricity at its best

We employ a unique infrared lens, invisible to the naked eye, that generates a motion signal 100 times stronger than standard motion sensors.

Beginning with the motion signal provides an intuitive gateway to unraveling the day's activities. The solitary dark blue segment indicates a shower, while the brown segment marks a bowel movement. From the various home sensors, our AI has identified periods of sleep, noting that the resident woke up twice to visit the bathroom.

Interestingly, you'll notice there's no motion detected during shower times because the monitored individual is blocking the sensor. This effect is also visible in the light intensity readings. Our AI picks up on such interactions. Concerns may be raised if this behavior changes, indicating a shift in the person's habits or attention.

Recorded values:  motion quantity, motion speed, motion direction, distance to sensor.

Sound frequencies

No voice recording is possible

We've intentionally limited our sound detector through hardware to exclude frequencies used for voice identification, ensuring that voice recording is not possible.

Indeed, the challenge lies in detecting speech without deciphering its content. Deliberately limiting detection capabilities is crucial for maintaining privacy. It's essential for users to trust that our sensors respect their privacy, ensuring they feel comfortable placing them in private spaces without hesitation.

Recorded values: sound level, 20 bands spectrogram, zero crossing rate, impulse, pitch, silence level.

Light spectrometer

100 kHz / 10 visible wavelengths

Our sensors are unique in the world, capable of distinguishing between sunlight and artificial light.

Light plays a crucial role in detecting human activities, and we pay meticulous attention to light calibration. Our technology can distinguish between the light emitted from a TV and a lamp, and even identify the type of lamp, whether it's fluorescent or LED.

The individual has obscured the sensor while showering, a detail that's evident in the motion signal. Our AI is designed to detect and interpret such behaviors.

Recorded values: intensity, 10 wavelengths intensity, 10 variation frequencies

Temperature

Extreme sensitivity 0.01ºC

With this level of precision, we're able to detect subtle events, such as water boiling or even the heat emitted by a refrigerator.

In fact, temperature detection posed the greatest challenge for our sensors. We've achieved very high sensitivity in detecting minor variations in air temperature. This enables us to identify the cycle of an air conditioning regulation device from up to 10 meters away. To attain such precision, we meticulously calibrate all our devices.

During shower time, the bathroom temperature increases by 4ºC. This showcases the reactivity and accuracy of our temperature measurement.

Recorded values: temperature

Water pressure

Measuring the precise number of water molecules in the air

We use a highly sensitive electrochemical sensor that catches water molecules through the porous surface of the sensor.

While humidity is a concept widely understood by the general public, it's not the most effective measure for detecting human activities. Partial pressure of water provides a more accurate assessment, as it solely depends on the amount of water molecules in the air and is independent of other environmental factors such as temperature and air pressure.

During the shower time, the vapor pressure in the bathroom rises significantly and remains quite stable for the rest of the day.

Recorded values: humidity, partial vapor pressure, dew point, air pressure, saturation vapor pressure.

CO2

High sensitivity 5ppm. Fast response time 10 ppm/s. Suited for human activities detection.

Unlike most CO2 sensors that require regular calibration in fresh air, ours do not. We've integrated an industrial-grade sensor into our device, boasting exceptionally high sensitivity.

Measuring CO2 levels is complex, involving an optical chamber for infrared emission, interferometric filters, and advanced signal processing. However, CO2 concentration is crucial for detecting human activities, as it can arise from breathing, bodily gases, or combustion processes like gas cooking. Our AI identifies the source of CO2 variations from the profile of the curve. Such characteristics are reached through a proprietary calibration process of our sensors.

Recorded values: CO2 concentration, air purity, traces of dust.

Indoor Air Velocity

Our exclusive sensing

Our device uses a proprietary sensor for measuring the velocity of the air over its sensitive surface.

Draughts, combined with CO2 concentration and temperature, serve as excellent indicators of active air quality management. Moreover, draughts are particularly crucial for the well-being of frail individuals.

Recorded values: air velocity, air pressure.