2.3 SOLI: A Tiny Device for a New Human Machine Interface

Trotta, S.; Weber, Dave; Jungmaier, Reinhard W.; Baheti, Ashutosh; Lien, Jaime; Noppeney, Dennis; Tabesh, Maryam; Rumpler, Christoph; Aichner, Michael; Albel, Siegfried; Bal, Jagjit S.; Poupyrev, Ivan

doi:10.1109/isscc42613.2021.9365835

Cited by 31 publications

(16 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this work, gesture sensing is performed by the BGT60TR13C FMCW radar sensor [44], manufactured by Infineon Technologies AG. The sensor is equipped with a Transmit (TX) and tree Receive (RX) channels with an included antenna integrated into the package.…”

Section: B Radar Boardmentioning

confidence: 99%

“…We first design a radar-based setup and preprocessing suitable for the meta-learning context. Using the sensor BGT60TR13C FMCW [44] we gather data for a total of twenty hand gestures performed by five users in three different environments. The collected raw data follow a definite frequency preprocessing and are then elaborated on the time axis for dimensionality reduction without relevant information loss.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Few-Shot User-Definable Radar-Based Hand Gesture Recognition at the Edge

et al. 2022

View full text Add to dashboard Cite

Technological advances and scalability are leading Human-Computer Interaction (HCI) to evolve towards intuitive forms, such as through gesture recognition. Among the various interaction strategies, radar-based recognition is emerging as a touchless, privacy-secure, and versatile solution in different environmental conditions. Classical radar-based gesture HCI solutions involve deep learning but require training on large and varied datasets to achieve robust prediction. Innovative self-learning algorithms can help tackling this problem by recognizing patterns and adapt from similar contexts. Yet, such approaches are often computationally expensive and hardly integrable into hardware-constrained solutions. In this paper, we present a gesture recognition algorithm which is easily adaptable to new users and contexts. We exploit an optimization-based meta-learning approach to enable gesture recognition in learning sequences. This method targets at learning the best possible initialization of the model parameters, simplifying training on new contexts when small amounts of data are available. The reduction in computational cost is achieved by processing the radar sensed data of gestures in the form of time maps, to minimize the input data size. This approach enables the adaptation of simple convolutional neural network (CNN) to new hand poses, thus easing the integration of the model into a hardware-constrained platform. Moreover, the use of a Variational Autoencoders (VAE) to reduce the gestures' dimensionality leads to a model size decrease of an order of magnitude and to half of the required adaptation time. The proposed framework, deployed on the Intel ® Neural Compute Stick 2 (NCS 2), leads to an average accuracy of around 84% for unseen gestures when only one example per class is utilized at training time. The accuracy increases up to 92.6% and 94.2% when three and five samples per class are used.

show abstract

Section: B Radar Boardmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Few-Shot User-Definable Radar-Based Hand Gesture Recognition at the Edge

et al. 2022

View full text Add to dashboard Cite

show abstract

“…The echo was recorded by the radar using a signal bandwidth B = 1 GHz corresponding to 15 cm range resolution, carrier frequency f c = 24 GHz, and a sweep repetition rate f PRF = 1 kHz. The waving hand, imitating a gesture used, for example, in the Google Soli radar [45], was located ca. 50 cm from the antenna aperture.…”

Section: Real-life Data Analysismentioning

confidence: 99%

Real-Time Variants of Vertical Synchrosqueezing: Application to Radar Remote Sensing

Abratkiewicz

Gambrych

2022

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

This paper presents thorough research on using high-order vertical synchrosqueezing (VSS) in different radar remote sensing applications. The method well-established in the literature is examined and compared to the novel form of third-order VSS and first-order VSS using an enhanced estimator of the instantaneous frequency (IF), both proposed by the authors. An investigation shows that the two introduced variants of the VSS are characterized by preserved capabilities (understood as the possibility to concentrate the time-frequency (TF) distribution and its reconstruction) with significantly reduced computation cost. The research shows that in practical radar remote sensing applications, the high-order VSS can be successfully replaced by the approach proposed in this paper with a lower computational burden. Furthermore, the methods are validated under numerical experiments, both simulated and real-life, which showed the efficiency of the proposed methods in radar signal processing, particular component extraction, and signal decomposition. Moreover, the authors developed the realtime graphics processing unit (GPU)-based implementation of the proposed techniques and presented its efficiency in practical conditions.

show abstract

“…The smallest model size has been achieved by the CNN2D classifier, where the sizes for the non-optimized and optimized versions are 375. 89 The last analyzed parameter is the mean inference time. This parameter strongly influences the interaction experience, which plays a particular role during real-time system operations.…”

Section: Comparison With Existing Techniquesmentioning

confidence: 99%

“…Figure 3. BGT60TR13C radar sensor block diagram[89]. The signal sensed by the three receiver channels (RX1, RX2, and RX3) is mixed with the transmitted signal from TX1, processed, and then converted digitally through the ADC.…”

mentioning

confidence: 99%

Highly-Optimized Radar-Based Gesture Recognition System with Depthwise Expansion Module

Chmurski

Mauro

Santra

et al. 2021

Sensors

View full text Add to dashboard Cite

The increasing integration of technology in our daily lives demands the development of more convenient human–computer interaction (HCI) methods. Most of the current hand-based HCI strategies exhibit various limitations, e.g., sensibility to variable lighting conditions and limitations on the operating environment. Further, the deployment of such systems is often not performed in resource-constrained contexts. Inspired by the MobileNetV1 deep learning network, this paper presents a novel hand gesture recognition system based on frequency-modulated continuous wave (FMCW) radar, exhibiting a higher recognition accuracy in comparison to the state-of-the-art systems. First of all, the paper introduces a method to simplify radar preprocessing while preserving the main information of the performed gestures. Then, a deep neural classifier with the novel Depthwise Expansion Module based on the depthwise separable convolutions is presented. The introduced classifier is optimized and deployed on the Coral Edge TPU board. The system defines and adopts eight different hand gestures performed by five users, offering a classification accuracy of 98.13% while operating in a low-power and resource-constrained environment.

show abstract

2.3 SOLI: A Tiny Device for a New Human Machine Interface

Cited by 31 publications

References 7 publications

Few-Shot User-Definable Radar-Based Hand Gesture Recognition at the Edge

Few-Shot User-Definable Radar-Based Hand Gesture Recognition at the Edge

Real-Time Variants of Vertical Synchrosqueezing: Application to Radar Remote Sensing

Highly-Optimized Radar-Based Gesture Recognition System with Depthwise Expansion Module

Contact Info

Product

Resources

About