MACHINE LEARNING / AI

Infinite Possibilities

 

What would you do if you could detect a person in a specific space at a specific time, only using your existing IoT devices? Turn the lights on for them? Adjust the heating and air conditioner settings? Notify the authorities?

Infinite Possibilities

What would you do if you could detect a person in a specific space at a specific time, only using your existing IoT devices? Turn the lights on for them? Adjust the heating and air conditioner settings? Notify the authorities?

Network Presence Sensing (NPS)

Ivani’s Network Presence Sensing transforms existing RF devices into wireless sensors without disrupting their primary application.

Network Presence Sensing (NPS)

Ivani’s Network Presence Sensing transforms existing RF devices into wireless sensors without disrupting their primary application.

True Occupancy Sensing

Unlike most “occupancy sensors” that only detect motion, NPS determines the presence and absence of people in a space, regardless of motion.

IVANI_overhead-view-of-busy-couple-in-modern-kitchen

Ivani’s occupancy sensing algorithms match the performance of current motion sensing hardware found in security hardware and commercial lighting controls, while also providing additional information. NPS goes beyond existing technology with the ability to detect humans even when they’re not in motion. It can also estimate the number of people in a space at any given time, whether they’re moving or not. How does it work? NPS uses an emergent system behavior of wireless communications to detect human activity or non-activity. The wireless network transmissions between IoT devices interact with the environment as they propagate through time and space. We then analyze the network diagnostic information to reveal human activity or non-activity in a detection area.

The challenge of human sensing is that the wireless transmissions within a space change with the room size and layout, the number of the people occupying the space, the physique of people in a space, and the level of activity in adjacent spaces (since RF can propagate through walls). The dimensionality of the data sets, the number of variables impacting the data, and the complexity of their interactions all contribute to the challenges associated with true presence sensing using commercial IoT devices.

Ivani_NPS_ML-AI_through walls-02-web

True Occupancy Sensing

Unlike most “occupancy sensors” that only detect motion, NPS determines the presence and absence of people in a space, regardless of motion.

IVANI_overhead-view-of-busy-couple-in-modern-kitchen

Ivani’s occupancy sensing algorithms match the performance of current motion sensing hardware found in security hardware and commercial lighting controls, while also providing additional information. NPS goes beyond existing technology with the ability to detect humans even when they’re not in motion. It can also estimate the number of people in a space at any given time, whether they’re moving or not. How does it work? NPS uses an emergent system behavior of wireless communications to detect human activity or non-activity. The wireless network transmissions between IoT devices interact with the environment as they propagate through time and space. We then analyze the network diagnostic information to reveal human activity or non-activity in a detection area.

Ivani_NPS_ML-AI_through walls-02-web

The challenge of human sensing is that the wireless transmissions within a space change with the room size and layout, the number of the people occupying the space, the physique of people in a space, and the level of activity in adjacent spaces (since RF can propagate through walls). The dimensionality of the data sets, the number of variables impacting the data, and the complexity of their interactions all contribute to the challenges associated with true presence sensing using commercial IoT devices.

NPS Artificial Intelligence

Despite these challenges, human activity sensing needs to operate under a variety of conditions and Ivani achieves this by leveraging innovative time-series analysis techniques coupled with cutting-edge artificial intelligence algorithms. These analyses and algorithms form the backbone of computational models that make predictions in real time and adapt to changes in the room’s environment.

All models undergo rigorous testing in Ivani’s comprehensive test facility. Each model’s performance is measured using quantitative data metrics that have been tailored to the end application. The test facility is “ground truthed” with an extensive video system to enable validated performance metrics with respect to both human activity sensing and the room’s environmental conditions.

Blue-Chip Data Engineers

Artificial intelligence is not a magic box, the “garbage in, garbage out” mantra still applies. The key to Ivani’s success are the minds behind extracting generalized knowledge from the data. Ivani’s team of data engineers have a knack for making stubborn datasets reveal their secrets by drawing on their backgrounds in signal processing, real-time control, deep learning, and human neurology. When traditional descriptive data analysis fails to provide insightful information, the Ivani engineers develop their own analysis to uncover information encoded in the RF environment.

This can be time-consuming and tedious, but it is critical to get right. Data sets engineered by the Ivani team feed the AI models and the models are only as good as the data used to build them.

NPS Artificial Intelligence

Despite these challenges, human activity sensing needs to operate under a variety of conditions and Ivani achieves this by leveraging innovative time-series analysis techniques coupled with cutting-edge artificial intelligence algorithms. These analyses and algorithms form the backbone of computational models that make predictions in real time and adapt to changes in the room’s environment.

All models undergo rigorous testing in Ivani’s comprehensive test facility. Each model’s performance is measured using quantitative data metrics that have been tailored to the end application. The test facility is “ground truthed” with an extensive video system to enable validated performance metrics with respect to both human activity sensing and the room’s environmental conditions.

Blue-Chip Data Engineers

Artificial intelligence is not a magic box, the “garbage in, garbage out” mantra still applies. The key to Ivani’s success are the minds behind extracting generalized knowledge from the data. Ivani’s team of data engineers have a knack for making stubborn datasets reveal their secrets by drawing on their backgrounds in signal processing, real-time control, deep learning, and human neurology. When traditional descriptive data analysis fails to provide insightful information, the Ivani engineers develop their own analysis to uncover information encoded in the RF environment.

This can be time-consuming and tedious, but it is critical to get right. Data sets engineered by the Ivani team feed the AI models and the models are only as good as the data used to build them.

Engineered Data and Predictive Models for Your Application

“One size fits all” often just isn’t good enough. Different applications require different types of engineered data. Whether they’re creating a new data class or putting a new spin on an old trick, Ivani’s engineering team puts the system first by leveraging their years of experience in developing application specific data classes. Requirements associated with a true occupancy sensing system often change with the application. For instance, the behavior of an HVAC system may be altered based on the estimated number of people occupying the space. Or lighting control applications, for example, typically require fast response times while HVAC control applications operate on a much slower time scale. If required response times are too fast it can possibly cause false presence predictions while slower response times allow decisions to be made with more confidence

Ivani_ML-AI_timing+accuracy-web2
IVANI_Abstract Background Lines And Dots

The consequences of inaccurate models change with the application. Security applications require accuracy in both presence and absence to ensure the system is trustworthy. On the other hand, false presence predictions in lighting applications might cause lights to turn on while no one is in the room, but it’s not acceptable if lighting applications produce false presence predictions that turn a light off while someone is in the room. It’s important to match your desired goals with the right application of occupancy sensing to provide end users with an optimal experience.

Ultimately, engineering the data and developing predictive models to fit the system requirements ensures that the delivered presence sensing system works as intended and elevates the overall user experience, because no one wants to be left in the dark.

Engineered Data and Predictive Models for Your Application

Ivani_ML-AI_timing+accuracy-web2

“One size fits all” often just isn’t good enough. Different applications require different types of engineered data. Whether they’re creating a new data class or putting a new spin on an old trick, Ivani’s engineering team puts the system first by leveraging their years of experience in developing application specific data classes. Requirements associated with a true occupancy sensing system often change with the application. For instance, the behavior of an HVAC system may be altered based on the estimated number of people occupying the space. Or lighting control applications, for example, typically require fast response times while HVAC control applications operate on a much slower time scale. If required response times are too fast it can possibly cause false presence predictions while slower response times allow decisions to be made with more confidence.

IVANI_Abstract Background Lines And Dots

The consequences of inaccurate models change with the application. Security applications require accuracy in both presence and absence to ensure the system is trustworthy. On the other hand, false presence predictions in lighting applications might cause lights to turn on while no one is in the room, but it’s not acceptable if lighting applications produce false presence predictions that turn a light off while someone is in the room. It’s important to match your desired goals with the right application of occupancy sensing to provide end users with an optimal experience.

Ultimately, engineering the data and developing predictive models to fit the system requirements ensures that the delivered presence sensing system works as intended and elevates the overall user experience, because no one wants to be left in the dark.

Ready to join the team?

We are building a world-class Machine Learning & AI team at Ivani.
Come join us so we can do great things together!

Ready to join the team?

We are building a world-class Machine Learning & AI team at Ivani.
Come join us so we can do great things together!

Ivani is a member of the

nvidia_white inception program - Ivani

Ivani is a member of the

nvidia_white inception program - Ivani