Browsing by Author "Sreekumar, S."

Filter results by typing the first few letters
Now showing 1 - 3 of 3
  • No Thumbnail Available
    Item
    Adult and Child Classification using Automotive Radar for In-cabin Monitoring
    (Institute of Electrical and Electronics Engineers Inc., 2024) Sreekumar, S.; Shashank, S.K.; Srihari, P.; Vandana, G.S.; Pardhasaradhi, B.; Cenkarmaddi, L.R.
    The awareness and decision-making about the unattended child or pet inside a car is one of the emerging features in autonomous vehicles as a precaution to prevent hot car death. The automotive radars can provide the Doppler and spatial information about in-cabin passengers. This paper proposes to extract the range-Doppler images from the IQ radar data and process them using CNNs to classify the passenger as an adult or child. The IWR1642 radar module is used to collect the passenger details in both space and time within the car. A novel CNN architecture is proposed by trading off the accuracy and lightweight characteristics of the network. The proposed architecture provides 97.74± 0.34 accuracy (with 18.32 MB size) compared to the denseNet201 of 99.13± 0.3 (with 71.3 MB size) accuracy. The proposed architecture is compared against the existing pre-trained models like InceptionNet, MobileNet, EfficientNet, NASNet, VGGNet, DenseNet, ResNet, and Xception regarding accuracy and size. © 2024 IEEE.
  • No Thumbnail Available
    Item
    Enhanced Intruder Detection Using mmWave Radar: A Spatiotemporal Clustering Approach for Robust Human Detection
    (Institute of Electrical and Electronics Engineers Inc., 2024) Shashank, S.K.; Sreekumar, S.; Manjith Srijan, M.; Srihari, P.; Vandana, G.S.; Pardhasaradhi, B.P.; Cenkarmaddi, L.R.
    Intruder detection using mmWave radar presents significant challenges due to the variability in the number of detections across spatial and temporal domains. This variability complicates the clustering process, particularly when detections from different body parts, such as the head and gait, are spatially distant, leading to the potential fragmentation of clusters. To address these challenges, we propose an enhanced clustering methodology that integrates both spatial and temporal information. The approach modifies the traditional DBSCAN algorithm by introducing a delayed window accumulation technique. This technique allows the radar system to accumulate detections over a specified duration, ensuring that an adequate number of detections are gathered before initiating clustering. Additionally, our method considers the physical dimensions of the human body to merge clusters that may be incorrectly separated due to the spatial distribution of detections. We further refine this approach by implementing a modified agglomerative clustering algorithm that leverages both spatial and temporal data to enhance cluster stability. The proposed methods are evaluated against the standard DBSCAN algorithm, demonstrating superior performance in accurately identifying human intruders, even under challenging detection scenarios. Our findings suggest that incorporating a delayed window accumulation strategy and considering spatiotemporal data are critical for robust intruder detection using mmWave radar. © 2024 IEEE.
  • No Thumbnail Available
    Item
    FMCW Radar-Based UAV Detection and Tracking Using Transfer Learning
    (Institute of Electrical and Electronics Engineers Inc., 2024) Sreekumar, S.; Shashank, S.K.; Srihari, P.; Nandagiri, A.; Vandana, G.S.; Pardhasaradhi, B.P.; Cenkarmaddi, L.R.
    This research investigation offers a novel method for monitoring and detecting unmanned aerial vehicles (UAVs) by combining transfer learning neural networks with Frequency Modulated Continuous Wave (FMCW) radar. The system utilizes a 60 GHz Texas Instruments IWR6843ISK radar with a DCA1000 board to capture raw radar signals, which are subsequently processed to generate range-angle heat maps. Ground truth data for UAV positioning is meticulously obtained using a dual GPS setup, where one GPS is stationed at the radar and the other is mounted on the UAV. The processed range-angle heat maps serve as the input for various transfer learning models, including DenseNet, InceptionV3, MobileNet, ResNet, and VGG, which are employed to compute the range data and angle data of the UAV. The results emphasize the potential of transfer learning in improving radar signal processing by demonstrating the effectiveness of these models in attaining accurate UAV detection and tracking. This approach is pivotal for applications requiring precise UAV monitoring, offering a robust solution for scenarios where traditional radar systems may fall short. The study underscores the advantages of leveraging transfer learning for improved radar-based UAV detection and sets the stage for future advancements in autonomous aerial monitoring and surveillance systems. © 2024 IEEE.