Compact highly-sensitive 140GHz MIMO radar system developed by imec

The MIMO radar system (multiple inputs with multiple inputs) of 140 GHz works up to a range of 10 m, with a resolution of 15 mm and a frequency of 10 GHz. Multiple antenna paths are built to allow a complete (virtual) 1×4 MIMO configuration to achieve angle separation of the target. Transmitter chip has chip antennas, and are integrated in 28nm bulk CMOS technology, providing a bargain solution for large volume production.

Adding machine learning opportunities, imec also demonstrated the ability of the radar to detect and classify small movements based on Doppler’s information.

“It opens up new opportunities, for example, allowing for recognition of gestures for intuitive human-machine interaction,” explains Barend van Liempd, head of research and development at imec. Consider AR / VR space, where the new radar can support intuitive interaction with virtual objects. Motion recognition can potentially enable intuitive device control – complementary to existing interfaces such as voice control or smart touch screens.

Since it is not sensitive to lighting conditions and keeps privacy (radar has not recognized people so far), radar solutions have special advantages over other types of motion sensors, such as infrared-based flight cameras. Being extremely compact, the 140 GHz radar system can invisiblely be integrated into virtually any device, such as laptops, smart phones, or screen frames.

The name has developed a specific machine learning algorithm based on a multilayer neuron network, including the LSTM layer and using controlled learning to train the capture model using internal shots of more than 25 people, including a few shots for each of the 7 different gestures. Against the experimental data set, the model classifies the recorded 7 gestures and predicts the real gesture at least 94% of the time.

Apart from the gesture, vital signs can be measured with a high precision thanks to the high radio frequency. Therefore, the radar is an excellent candidate for vital vehicle monitoring systems to allow for contactless tracking of driver states, eg detecting fallopnea, abnormal levels of stress or possibly preventing accidents due to acute health hazards such as heart attacks or epilepsy. Another possible application is to track children by detecting movement and vital signs, even when the child is covered with a blanket and a sleep, eg providing a warning if the child is unintentionally left in the vehicle.

In order to increase the wealth of data and spatial information, imec is currently building a 4×4 MIMO radar system, which is developing a new generation of radar chips – including TX and RX as separate chips. This will provide greater flexibility in the distribution of MIMO array elements in the available area.