Research Interests
On-device LLM
Computer Vision
Automated Speech Recognition
Natural Language Processing,
Multimodal Foundation Model
Artificial Intelligence of Things (AIoT)
Wireless Communication
Robotics and IoT
Publications
Title : "Artificial Intelligence of Things: A Survey"
Authors: Shakhrul Iman Siam, Hyunho Ahn, Li Liu, Samiul Alam, Hui Shen, Zhichao Cao, Ness Shroff, Bhaskar Krishnamachari, Mani Srivastava, Mi Zhang
Journal: ACM Transactions on Sensor Networks (ACM TOSN)
Abstract:
The integration of the Internet of Things (IoT) and modern Artificial Intelligence (AI) has given rise to a new paradigm known as the Artificial Intelligence of Things (AIoT). In this survey, we provide a systematic and comprehensive review of AIoT research. We examine AIoT literature related to sensing, computing, and networking & communication, which form the three key components of AIoT. In addition to advancements in these areas, we review domain-specific AIoT systems that are designed for various important application domains. We have also created an accompanying GitHub repository, where we compile the papers included in this survey: https://github.com/AIoT-MLSys-Lab/AIoT-Survey. This repository will be actively maintained and updated with new research as it becomes available. As both IoT and AI become increasingly critical to our society, we believe AIoT is emerging as an essential research field at the intersection of IoT and modern AI. We hope this survey will serve as a valuable resource for those engaged in AIoT research and act as a catalyst for future explorations to bridge gaps and drive advancements in this exciting field.
Title : "ChirpTransformer: Versatile LoRa Encoding for Low-power Wide-area IoT"
Authors: Chenning Li, Yidong Ren, Shuai Tong, Shakhrul Iman Siam, Mi Zhang, Jiliang Wang, Yunhao Liu, Zhichao Cao
Conference: Proceedings of the 22nd Annual International Conference on Mobile Systems, Applications and Services (Mobisys 24)
Abstract:
This paper introduces ChirpTransformer, a versatile LoRa encoding framework that harnesses broad chirp features to dynamically modulate data, enhancing network coverage, throughput, and energy efficiency. Unlike the standard LoRa encoder that offers only single configurable chirp feature, our framework introduces four distinct chirp features, expanding the spectrum of methods available for data modulation. To implement these features on commercial off-the-shelf (COTS) LoRa nodes, we utilize a combination of a software design and a hardware interrupt. ChirpTransformer serves as the foundation for optimizing encoding and decoding in three specific case studies: weak signal decoding for extended network coverage, concurrent transmission for heightened network throughput, and data rate adaptation for improved network energy efficiency. Each case study involves the development of an end-to-end system to comprehensively evaluate its performance. The evaluation results demonstrate remarkable enhancements compared to the standard LoRa. Specifically, ChirpTransformer achieves a 2.38 × increase in network coverage, a 3.14 × boost in network throughput, and a 3.93 × of battery lifetime.
Title : "A Deep Learning Based Person Detection and Heatmap Generation Technique with a Multi-Camera System"
Authors: Md. Shakhrul Iman Siam, Subrata Biswas
Keywords: Object Detection, Heatmap, KDE, Homography Transform
Conference: 12th International Conference On Electrical and Computer Engineering, 2022 (ICECE)
Abstract:
This paper outlines a technical method for video analysis that may be used to identify persons in footage from several CCTV cameras and provide a heatmap of that information for a certain floor layout. The analysis of customer and employee behavior in retail and office settings, as well as motion tracking and advertising effectiveness research, can all be aided by the automatic creation of people density maps. With the use of video recordings made by common video surveillance cameras, density maps were created. We made advantage of CCTV cameras, which are dispersed across a retail establishment. Because the Yolov5 object detection algorithm may produce findings more quickly, we have chosen to employ it for human detection. Additionally, due to the short inference time, it is appropriate for real-time applications.
Title : "BRL Signal Based Multi-Class Sleep Stage Classification using Feature Extraction and Random Forest Classifier"
Authors: Md. Shakhrul Iman Siam, Saiful Bari Siddiki, Mushfiqul Abedin, Dr. Muhammed Imamul Hassan Bhuiyan.
Conference: 12th International Conference On Electrical and Computer Engineering, 2022 (ICECE)
Abstract:
Sleep disorders are a common problem that disrupts our regular sleeping patterns. To diagnose sleep disorders, Long-term monitoring of sleep could be useful. In this paper an automated scheme of sleep staging is presented based on Bioradiolocation signals using time and frequency domain feature extraction and Random Forest Classifier. This experiment is val idated using data of 32 subjects without sleep-related breathing disorders. A Random Forest based algorithm is used for two, three, four and five-stage classification. We achieved the best performance so far (89.35% accuracy and 0.65 Cohens kappa) on 2-stage, 75.3% accuracy on 3-stage, 56.18% on 4-stage, and 54.2% accuracy on 5-stage classification with BRL Signals. These results show high potential in real-life sleep stage monitoring systems.
Title: "A Dual Purpose Refreshable Braille Display Based On Real Time Object Detection And Optical Character Recognition"
Authors: K M Naimul Hasan, Subrata Kumar Biswas, Md. Shakil Anwar, Md. Shakhrul Iman Siam & Celia Shahnaz
Conference: 2019 IEEE International Conference on Signal Processing, Information, Communication & Systems (SPICSCON)
Abstract:
This paper proposes a dual-purpose braille system for the visually impaired people. There are two main features of this system- object detection and optical character recognition. Real time object detection will help a visually impaired person to know about the things around him and optical character recognition will help him reading characters in both international (English) and local community (Bengali) language. In this paper, the detailed methodology of our proposed method is described. A pre-trained convolutional neural network (AlexNet) is used for classifying the objects and an OCR engine (Tesseract) along with basic image processing is used for optical character recognition. A refreshable braille display is also designed to show the braille characters.
Developed Projects
real-time video analysis and insight generation from multiple camera networks in a retail outlet.
Full pipeline of a Facial recognition-based automated attendance system.
Person Detection using YOLOv5 model trained on a custom dataset, and Re-identification using a Deep Learning model and customized algorithms. This can be used to track people's movement and activities in a Retail store or any place that is covered with CCTV cameras.
An Optical Character Recognition (OCR) based application that can extract Medicine Names and other information from a Hand-written prescription.
Generate Synthetic Handwritten images of various font from input Text
Prototype of a low-cost mobile device, that can assist a visually disabled person to read, recognize person, and detect objects using voice command. There are three main features of this system- Object detection, Face Recognition, and optical character recognition. Real-time object detection will help a visually impaired person to know about the things around him and optical character recognition will help him read characters in both international (English) and local community (Bengali) languages.
Competition Projects
IEEE Video and Image Processing CUP 2020
YESIST INNOVATION Challenge 2019
Other Projects
Simulating SLAM and LIDAR sensor on a 2D Floor plan using python.
It can take AC input and provide variable DC output adjusted by Potentiometer.
Other Research Works
Extracting SpO2 by pulse oximetry , X-Ray image processing, along with combining clinical symptoms and Machine Learning to detect Pneumonia using non-invasive method
Duration: August 2020 to December 2020
Supervisor: Dr. Celia Shahnaz
Keywords: CNN, Non-Invasive, SpO2
Description: Pneumonia is one of the most common infectious diseases in the lungs especially attacked infant and elderly people which causing cough with phlegm or pus, fever, chills, difficulty breathing, fill lungs with pus, and sometimes leads to deaths in some case. In Bangladesh, pneumonia is responsible for around 28% of the deaths of children under five years of age. Around 50,000 children die of pneumonia every year because of not having proper medical doctors, especially in rural areas. But early detection can help to prevent this disease. Developing a device to detect pneumonia in a rapid and non-invasive way, which will be affordable for Bangladeshi people is our aim. Helping medical professionals to detect pneumonia early to facilitate the treatment of the patients is our motivation. That why an affordable, rapid detection model is designed. In our device, we used machine learning techniques using a non-invasive technique method by collecting x-ray images, spo2 level along with clinical symptoms that can help to detect pneumonia in absence of a doctor. Moreover, only an x-ray or only spo2 can not specify a pneumonia patient. But our model takes all the info and is gathered in machine learning model to detect the chance of having pneumonia with great accuracy with respect to other proposed models.