This article introduces a time-domain-based artificial intelligence (AI) radar system for gesture recognition using 33-GS/s direct sampling technique. High-speed sampling using a time-extension method allows AI learning to be applied to a time-domain radar signal reflecting information on both dynamic and static gestures, and thus can recognize not only dynamic but also static gestures. The Vernier clock generators and high-speed active samplers applied with the time-extension technique makes sampling at 33 GS/s possible. A 1-D convolutional neural network and long short-term memory are employed for both static and dynamic gestures and recognition rates of 93.2% and 90.5% are obtained, respectively. The radar system is implemented using a 65-nm CMOS process with a power consumption of 95 mW.
- Artificial intelligence (AI) radar,
- gesture recognition,
- high-speed sampling,
- impulse radar ultra-wideband (IR-UWB),
- time-to-digital converter (TDC) ,
- wireless sensing
- Artificial intelligence,
- Time-domain analysis,
Conventional radar systems have been used to detect the range, angle, and velocity of objects , . However, recently, a learning-based radar system using artificial intelligence (AI) technology has emerged and has begun to be used for target recognition . Such an advanced radar system can process significantly more information than a conventional radar approach and numerous studies are being conducted to enable its use in various applications. In particular, in –, a new attempt at applying the AI radar to a hand-gesture recognition system is described.
In , and AI radar system for gesture recognition is implemented by applying machine learning to a conventional frequency-modulated continuous-wave (FMCW) radar. There are three main phases during object recognition: a receiving step, a data processing step, and a recognition step. During the receiving process, a system board receives the reflected continuous wave from the target. The received signal includes the distinguishable Doppler shifts based on object movement. In addition, the differences in Doppler shift are visualized through the range-Doppler image (RDI) during the data processing stage . Finally, during the recognition phase, the AI radar system learns and recognizes the features of the RDI generated according to the gesture movements.
However, the AI radar system described in  using the Doppler effect is difficult to recognize static gestures. (Because static gestures do not generate Doppler shifts.) By contrast, when analyzing signals by sampling at a high speed within the time domain, information including the position and the shape of the static gesture can be obtained. In addition, it is possible to recognize static gestures through a learning process. Moreover, if the system considers a series of static gestures as a single pattern, it can recognize even dynamic gestures. Therefore, in this article, a time-domain AI radar system that recognizes both static and dynamic gestures is proposed .
Fig. 1 shows a conceptual diagram of the proposed time-domain AI radar system. The impulse signals reflected from the target have different waveforms according to the position and shape of the target. This reflects the inherent characteristics of the gestures, as shown in the waveforms in Fig. 1. Gestures can be recognized by training a convolutional neural network (CNN)  or long short-term memory (LSTM)  with waveforms that include different characteristics of objects. The system can recognize an object even if there is no movement because the waveform analysis is conducted in the time domain without using the Doppler effect .
In this article, a time-domain based AI radar system using direct sampling at 33 GS/s is proposed. This system can recognize both static and dynamic gestures by learning the characteristics of the waveform returning from the target. High-speed sampling was processed using a Vernier clock generator and an active sampler structure. In addition, the high-speed sampled signal was converted into a low-speed signal through a time-extension method such that digital data can be generated using a low-speed, low-power ADC. By applying 1-D-CNN and LSTM, recognition rates of 93.2% and 90.5% were recorded for five types of static gestures and six types of dynamic gestures, respectively.
The authors would like to thank Rohde and Schwarz for their support of the test instruments applied.
The Kavian Scientific Research Association (KSRA) is a non-profit research organization to provide research / educational services in December 2013. The members of the community had formed a virtual group on the Viber social network. The core of the Kavian Scientific Association was formed with these members as founders. These individuals, led by Professor Siavosh Kaviani, decided to launch a scientific / research association with an emphasis on education.
KSRA research association, as a non-profit research firm, is committed to providing research services in the field of knowledge. The main beneficiaries of this association are public or private knowledge-based companies, students, researchers, researchers, professors, universities, and industrial and semi-industrial centers around the world.
Our main services Based on Education for all Spectrum people in the world. We want to make an integration between researches and educations. We believe education is the main right of Human beings. So our services should be concentrated on inclusive education.
The KSRA team partners with local under-served communities around the world to improve the access to and quality of knowledge based on education, amplify and augment learning programs where they exist, and create new opportunities for e-learning where traditional education systems are lacking or non-existent.
FULL Paper PDF file:A Time Domain Artificial Intelligence Radar System Using 33-GHz Direct Sampling for Hand Gesture Recognition
A Time-Domain Artificial Intelligence Radar System Using 33-GHz Direct Sampling for Hand Gesture Recognition
in the IEEE Journal of Solid-State Circuits, vol. 55, no. 4, pp. 879-888, April 2020
PDF reference and original file: Click here
Professor Siavosh Kaviani was born in 1961 in Tehran. He had a professorship. He holds a Ph.D. in Software Engineering from the QL University of Software Development Methodology and an honorary Ph.D. from the University of Chelsea.