Showing 1–21 of 21 results for author: Roy, T

Koopman Operator-based Detection-Isolation of Cyberattack: A Case Study on Electric Vehicle Charging

Authors: Sanchita Ghosh, Tanushree Roy

Abstract: One of the key challenges towards the reliable operation of cyber-physical systems (CPS) is the threat of cyberattacks on system actuation signals and measurements. In recent years, system theoretic research has focused on effectively detecting and isolating these cyberattacks to ensure proper restorative measures. Although both model-based and model-free approaches have been used in this context,… ▽ More One of the key challenges towards the reliable operation of cyber-physical systems (CPS) is the threat of cyberattacks on system actuation signals and measurements. In recent years, system theoretic research has focused on effectively detecting and isolating these cyberattacks to ensure proper restorative measures. Although both model-based and model-free approaches have been used in this context, the latter are increasingly becoming more popular as complexities and model uncertainties in CPS increases. Thus, in this paper we propose a Koopman operator-based model-free cyberattack detection-isolation scheme for CPS. The algorithm uses limited system measurements for its training and generates real-time detection-isolation flags. Furthermore, we present a simulation case study to detect and isolate actuation and sensor attacks in a Lithium-ion battery system of a plug-in electric vehicle during charging. △ Less

Submitted 24 June, 2024; originally announced June 2024.

Comments: 10 pages, 4 figures, to be published in 2024 American Control Conference

Deep Learning Based Joint Multi-User MISO Power Allocation and Beamforming Design

Authors: Cemil Vahapoglu, Timothy J. O'Shea, Tamoghna Roy, Sennur Ulukus

Abstract: The evolution of fifth generation (5G) wireless communication networks has led to an increased need for wireless resource management solutions that provide higher data rates, wide coverage, low latency, and power efficiency. Yet, many of existing traditional approaches remain non-practical due to computational limitations, and unrealistic presumptions of static network conditions and algorithm ini… ▽ More The evolution of fifth generation (5G) wireless communication networks has led to an increased need for wireless resource management solutions that provide higher data rates, wide coverage, low latency, and power efficiency. Yet, many of existing traditional approaches remain non-practical due to computational limitations, and unrealistic presumptions of static network conditions and algorithm initialization dependencies. This creates an important gap between theoretical analysis and real-time processing of algorithms. To bridge this gap, deep learning based techniques offer promising solutions with their representational capabilities for universal function approximation. We propose a novel unsupervised deep learning based joint power allocation and beamforming design for multi-user multiple-input single-output (MU-MISO) system. The objective is to enhance the spectral efficiency by maximizing the sum-rate with the proposed joint design framework, NNBF-P while also offering computationally efficient solution in contrast to conventional approaches. We conduct experiments for diverse settings to compare the performance of NNBF-P with zero-forcing beamforming (ZFBF), minimum mean square error (MMSE) beamforming, and NNBF, which is also our deep learning based beamforming design without joint power allocation scheme. Experiment results demonstrate the superiority of NNBF-P compared to ZFBF, and MMSE while NNBF can have lower performances than MMSE and ZFBF in some experiment settings. It can also demonstrate the effectiveness of joint design framework with respect to NNBF. △ Less

Submitted 12 June, 2024; originally announced June 2024.

Diagnosis Of Takotsubo Syndrome By Robust Feature Selection From The Complex Latent Space Of DL-based Segmentation Network

Authors: Fahim Ahmed Zaman, Wahidul Alam, Tarun Kanti Roy, Amanda Chang, Kan Liu, Xiaodong Wu

Abstract: Researchers have shown significant correlations among segmented objects in various medical imaging modalities and disease related pathologies. Several studies showed that using hand crafted features for disease prediction neglects the immense possibility to use latent features from deep learning (DL) models which may reduce the overall accuracy of differential diagnosis. However, directly using cl… ▽ More Researchers have shown significant correlations among segmented objects in various medical imaging modalities and disease related pathologies. Several studies showed that using hand crafted features for disease prediction neglects the immense possibility to use latent features from deep learning (DL) models which may reduce the overall accuracy of differential diagnosis. However, directly using classification or segmentation models on medical to learn latent features opt out robust feature selection and may lead to overfitting. To fill this gap, we propose a novel feature selection technique using the latent space of a segmentation model that can aid diagnosis. We evaluated our method in differentiating a rare cardiac disease: Takotsubo Syndrome (TTS) from the ST elevation myocardial infarction (STEMI) using echocardiogram videos (echo). TTS can mimic clinical features of STEMI in echo and extremely hard to distinguish. Our approach shows promising results in differential diagnosis of TTS with 82% diagnosis accuracy beating the previous state-of-the-art (SOTA) approach. Moreover, the robust feature selection technique using LASSO algorithm shows great potential in reducing the redundant features and creates a robust pipeline for short- and long-term disease prognoses in the downstream analysis. △ Less

Submitted 18 January, 2024; v1 submitted 19 December, 2023; originally announced December 2023.

Comments: 5 pages, 3 figures, conference

On Distinguishability of Anomalies as Physical Faults or Actuation Cyberattacks

Authors: Tanushree Roy, Satadru Dey

Abstract: Increased automation has created an impetus to integrate infrastructure with wide-spread connectivity in order to improve efficiency, sustainability, autonomy, and security. Nonetheless, this reliance on connectivity and the inevitability of complexity in this system increases the vulnerabilities to physical faults or degradation and external cyber-threats. However, strategies to counteract faults… ▽ More Increased automation has created an impetus to integrate infrastructure with wide-spread connectivity in order to improve efficiency, sustainability, autonomy, and security. Nonetheless, this reliance on connectivity and the inevitability of complexity in this system increases the vulnerabilities to physical faults or degradation and external cyber-threats. However, strategies to counteract faults and cyberattacks would be widely different and thus it is vital to not only detect but also to identify the nature of the anomaly that is present in these systems. In this work, we propose a mathematical framework to distinguish between physical faults and cyberattack using a sliding mode based unknown input observer. Finally, we present simulation case studies to distinguish between physical faults and cyberattacks using the proposed Distinguishability metric and criterion. The simulation results show that the proposed framework successfully distinguishes between faults and cyberattacks. △ Less

Submitted 15 November, 2023; originally announced November 2023.

Comments: This paper is currently submitted to ASME

Deep Learning Based Uplink Multi-User SIMO Beamforming Design

Authors: Cemil Vahapoglu, Timothy J. O'Shea, Tamoghna Roy, Sennur Ulukus

Abstract: The advancement of fifth generation (5G) wireless communication networks has created a greater demand for wireless resource management solutions that offer high data rates, extensive coverage, minimal latency and energy-efficient performance. Nonetheless, traditional approaches have shortcomings when it comes to computational complexity and their ability to adapt to dynamic conditions, creating a… ▽ More The advancement of fifth generation (5G) wireless communication networks has created a greater demand for wireless resource management solutions that offer high data rates, extensive coverage, minimal latency and energy-efficient performance. Nonetheless, traditional approaches have shortcomings when it comes to computational complexity and their ability to adapt to dynamic conditions, creating a gap between theoretical analysis and the practical execution of algorithmic solutions for managing wireless resources. Deep learning-based techniques offer promising solutions for bridging this gap with their substantial representation capabilities. We propose a novel unsupervised deep learning framework, which is called NNBF, for the design of uplink receive multi-user single input multiple output (MU-SIMO) beamforming. The primary objective is to enhance the throughput by focusing on maximizing the sum-rate while also offering computationally efficient solution, in contrast to established conventional methods. We conduct experiments for several antenna configurations. Our experimental results demonstrate that NNBF exhibits superior performance compared to our baseline methods, namely, zero-forcing beamforming (ZFBF) and minimum mean square error (MMSE) equalizer. Additionally, NNBF is scalable to the number of single-antenna user equipments (UEs) while baseline methods have significant computational burden due to matrix pseudo-inverse operation. △ Less

Submitted 28 September, 2023; originally announced September 2023.

A Cyberattack Detection-Isolation Scheme For CAV Under Changing Driving Environment

Authors: Sanchita Ghosh, Nutan Saha, Tanushree Roy

Abstract: Under a changing driving environment, a Connected Autonomous Vehicle (CAV) platoon relies strongly on the acquisition of accurate traffic information from neighboring vehicles as well as reliable commands from a centralized supervisory controller through the communication network. Even though such modalities are imperative to ensure the safe and efficient driving performance of CAVs, they led to m… ▽ More Under a changing driving environment, a Connected Autonomous Vehicle (CAV) platoon relies strongly on the acquisition of accurate traffic information from neighboring vehicles as well as reliable commands from a centralized supervisory controller through the communication network. Even though such modalities are imperative to ensure the safe and efficient driving performance of CAVs, they led to multiple security challenges. Thus, a cyberattack on this network can corrupt vehicle-to-infrastructure (V2I) and vehicle-to-vehicle (V2V) communication, which can lead to unsafe or undesired driving scenarios. Hence, in this paper, we propose a unified V2V and V2I cyberattack detection scheme along with a V2I isolation scheme for CAVs under changing driving conditions. The proposed scheme is constructed using a bank of residual generators with Lyapunov function-based performance guarantees, such as disturbance-to-state stability, robustness, and sensitivity. Finally, we showcase the efficacy of our proposed algorithm through two simulation studies, namely under highway driving and urban driving. The results show that the proposed scheme can enhance the cybersecurity of CAVs by detecting and isolating cyberattacks on CAV platoons. △ Less

Submitted 18 May, 2023; originally announced May 2023.

Comments: This work has been submitted to IEEE for review

Histopathological Cancer Detection Using Hybrid Quantum Computing

Authors: Reek Majumdar, Biswaraj Baral, Bhavika Bhalgamiya, Taposh Dutta Roy

Abstract: We present an effective application of quantum machine learning in the field of healthcare. The study here emphasizes on a classification problem of a histopathological cancer detection using quantum transfer learning. Rather than using single transfer learning model, the work model presented here consists of multiple transfer learning models especially ResNet18, VGG-16, Inception-v3, AlexNet and… ▽ More We present an effective application of quantum machine learning in the field of healthcare. The study here emphasizes on a classification problem of a histopathological cancer detection using quantum transfer learning. Rather than using single transfer learning model, the work model presented here consists of multiple transfer learning models especially ResNet18, VGG-16, Inception-v3, AlexNet and several variational quantum circuits (VQC) with high expressibility. As a result, we provide a comparative analysis of the models and the best performing transfer learning model with the prediction AUC of approximately 93 percent for histopathological cancer detection. We also observed that for 1000 images with Resnet18, Hybrid Quantum and Classical (HQC) provided a slightly better accuracy of 88.5 percent than classical of 88.0 percent. △ Less

Submitted 6 February, 2023; originally announced February 2023.

COVID-19 Detection using Transfer Learning with Convolutional Neural Network

Authors: Pramit Dutta, Tanny Roy, Nafisa Anjum

Abstract: The Novel Coronavirus disease 2019 (COVID-19) is a fatal infectious disease, first recognized in December 2019 in Wuhan, Hubei, China, and has gone on an epidemic situation. Under these circumstances, it became more important to detect COVID-19 in infected people. Nowadays, the testing kits are gradually lessening in number compared to the number of infected population. Under recent prevailing con… ▽ More The Novel Coronavirus disease 2019 (COVID-19) is a fatal infectious disease, first recognized in December 2019 in Wuhan, Hubei, China, and has gone on an epidemic situation. Under these circumstances, it became more important to detect COVID-19 in infected people. Nowadays, the testing kits are gradually lessening in number compared to the number of infected population. Under recent prevailing conditions, the diagnosis of lung disease by analyzing chest CT (Computed Tomography) images has become an important tool for both diagnosis and prophecy of COVID-19 patients. In this study, a Transfer learning strategy (CNN) for detecting COVID-19 infection from CT images has been proposed. In the proposed model, a multilayer Convolutional neural network (CNN) with Transfer learning model Inception V3 has been designed. Similar to CNN, it uses convolution and pooling to extract features, but this transfer learning model contains weights of dataset Imagenet. Thus it can detect features very effectively which gives it an upper hand for achieving better accuracy. △ Less

Submitted 17 June, 2022; originally announced June 2022.

Comments: 4 pages, 4 figures, 2nd International Conference on Robotics, Electrical and Signal Processing Techniques (ICREST), DHAKA, Bangladesh

Journal ref: 2nd International Conference on Robotics, Electrical and Signal Processing Techniques (ICREST), DHAKA, Bangladesh, 2021, pp. 429-432

An Input-to-State Safety Approach Towards Safe Control of a Class of Parabolic PDEs Under Disturbances

Authors: Tanushree Roy, Ashley Knichel, Satadru Dey

Abstract: Distributed Parameter Systems (DPSs), modelled by partial differential equations (PDEs), are increasingly vulnerable to disturbances arising from various sources. Although detection of disturbances in PDE systems have received considerable attention in existing literature, safety control of PDEs under disturbances remains significantly under-explored. In this context, we explore a practical input-… ▽ More Distributed Parameter Systems (DPSs), modelled by partial differential equations (PDEs), are increasingly vulnerable to disturbances arising from various sources. Although detection of disturbances in PDE systems have received considerable attention in existing literature, safety control of PDEs under disturbances remains significantly under-explored. In this context, we explore a practical input-to-state safety (pISSf) based control design approach for a class of DPSs modelled by linear Parabolic PDEs. Specifically, we develop a control design framework for this class of system with both safety and stability guarantees based on control Lyapunov functional and control barrier functional. To illustrate our methodology, we apply our strategy to design a thermal control system for battery modules under disturbance. Several simulation studies are done to show the efficacy of our method. △ Less

Submitted 20 June, 2023; v1 submitted 6 January, 2022; originally announced January 2022.

Comments: Submitted to IEEE for review

Journal ref: IEEE Transactions on Control Systems Technology, 2024

Wideband Signal Localization with Spectral Segmentation

Authors: Nathan West, Tamoghna Roy, Timothy O'Shea

Abstract: Signal localization is a spectrum sensing problem that jointly detects the presence of a signal and estimates a center frequency and bandwidth. This is a step beyond most spectrum sensing work which estimates "present" or "not present" detections for either a single channel or fixed sized channels. We define the signal localization task, present the metrics of precision and recall, and establish b… ▽ More Signal localization is a spectrum sensing problem that jointly detects the presence of a signal and estimates a center frequency and bandwidth. This is a step beyond most spectrum sensing work which estimates "present" or "not present" detections for either a single channel or fixed sized channels. We define the signal localization task, present the metrics of precision and recall, and establish baselines for traditional energy detection on this task. We introduce a new dataset that is useful for training neural networks to perform this task and show a training framework to train signal detectors to achieve the task and present precision and recall curves over SNR. This neural network based approach shows an 8 dB improvement in recall over the traditional energy detection approach with minor improvements in precision. △ Less

Submitted 1 October, 2021; originally announced October 2021.

Comments: arXiv admin note: substantial text overlap with arXiv:2110.00518

Journal ref: IEEE Asilomar 2021

A Wideband Signal Recognition Dataset

Authors: Nathan West, Timothy O'Shea, Tamoghna Roy

Abstract: Signal recognition is a spectrum sensing problem that jointly requires detection, localization in time and frequency, and classification. This is a step beyond most spectrum sensing work which involves signal detection to estimate "present" or "not present" detections for either a single channel or fixed sized channels or classification which assumes a signal is present. We define the signal recog… ▽ More Signal recognition is a spectrum sensing problem that jointly requires detection, localization in time and frequency, and classification. This is a step beyond most spectrum sensing work which involves signal detection to estimate "present" or "not present" detections for either a single channel or fixed sized channels or classification which assumes a signal is present. We define the signal recognition task, present the metrics of precision and recall to the RF domain, and review recent machine-learning based approaches to this problem. We introduce a new dataset that is useful for training neural networks to perform these tasks and show a training framework to train wideband signal recognizers. △ Less

Submitted 1 October, 2021; originally announced October 2021.

Journal ref: IEEE SPAWC 2021

Security of Distributed Parameter Cyber-Physical Systems: Cyber-Attack Detection in Linear Parabolic PDEs

Authors: Tanushree Roy, Satadru Dey

Abstract: Security of Distributed Parameter Cyber-Physical Systems (DPCPSs) is of critical importance in the face of cyber-attack threats. Although security aspects of Cyber-Physical Systems (CPSs) modelled by Ordinary differential Equations (ODEs) have been extensively explored during the past decade, security of DPCPSs has not received its due attention despite its safety-critical nature. In this work, we… ▽ More Security of Distributed Parameter Cyber-Physical Systems (DPCPSs) is of critical importance in the face of cyber-attack threats. Although security aspects of Cyber-Physical Systems (CPSs) modelled by Ordinary differential Equations (ODEs) have been extensively explored during the past decade, security of DPCPSs has not received its due attention despite its safety-critical nature. In this work, we explore the security aspects of DPCPSs from a system theoretic viewpoint. Specifically, we focus on DPCPSs modelled by linear parabolic Partial Differential Equations (PDEs) subject to cyber-attacks in actuation channel. First, we explore the detectability of such attacks and derive conditions for stealthy attacks. Next, we develop a design framework for cyber-attack detection algorithms based on output injection observers. Such attack detection algorithms explicitly consider stability, robustness and attack sensitivity in their design. Finally, theoretical analysis and simulation studies are performed to illustrate the effectiveness of the proposed approach. △ Less

Submitted 30 July, 2021; v1 submitted 29 July, 2021; originally announced July 2021.

Comments: This work has been submitted for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible

Journal ref: IEEE Transactions on Control Systems Technology, vol. 31, no. 6, pp. 2437-2448, Nov. 2023

Thermal Fault Detection and Localization Framework for Large Format Batteries

Authors: Sara Sattarzadeh, Tanushree Roy, Satadru Dey

Abstract: Safety against thermal failures is crucial in battery systems. Real-time thermal diagnostics can be a key enabler of such safer batteries. Thermal fault diagnostics in large format pouch or prismatic cells pose additional challenges compared to cylindrical cells. These challenges arise from the fact that the temperature distribution in large format cells is at least two-dimensional in nature (alon… ▽ More Safety against thermal failures is crucial in battery systems. Real-time thermal diagnostics can be a key enabler of such safer batteries. Thermal fault diagnostics in large format pouch or prismatic cells pose additional challenges compared to cylindrical cells. These challenges arise from the fact that the temperature distribution in large format cells is at least two-dimensional in nature (along length and breadth) while such distribution can be reasonably approximated in one dimension (along radial direction) in cylindrical cells. This difference makes the placement of temperature sensor(s) non-trivial and the design of detection algorithm challenging. In this work, we address these issues by proposing a framework that (i) optimizes the sensor locations to improve detectability and isolability of thermal faults, and (ii) designs a filtering scheme for fault detection and localization based on a two-dimensional thermal model. The proposed framework is illustrated by experimental and simulation studies on a commercial battery cell. △ Less

Submitted 12 September, 2021; v1 submitted 25 March, 2021; originally announced March 2021.

Journal ref: Journal of Power Sources, Volume 512, 15 November 2021, 230400

Cyber-Attack Detection in Socio-Technical Transportation Systems Exploiting Redundancies Between Physical and Social Data

Authors: Tanushree Roy, Sara Sattarzadeh, Satadru Dey

Abstract: Cyber-physical-social connectivity is a key element in Intelligent Transportation Systems (ITSs) due to the ever-increasing interaction between human users and technological systems. Such connectivity translates the ITSs into dynamical systems of socio-technical nature. Exploiting this socio-technical feature to our advantage, we propose a cyber-attack detection scheme for ITSs that focuses on cyb… ▽ More Cyber-physical-social connectivity is a key element in Intelligent Transportation Systems (ITSs) due to the ever-increasing interaction between human users and technological systems. Such connectivity translates the ITSs into dynamical systems of socio-technical nature. Exploiting this socio-technical feature to our advantage, we propose a cyber-attack detection scheme for ITSs that focuses on cyber-attacks on freeway traffic infrastructure. The proposed scheme combines two parallel macroscopic traffic model-based Partial Differential Equation (PDE) filters whose output residuals are compared to make decision on attack occurrences. One of the filters utilizes physical (vehicle/infrastructure) sensor data as feedback whereas the other utilizes social data from human users' mobile devices as feedback. The Social Data-based Filter is aided by a fake data isolator and a social signal processor that translates the social information into usable feedback signals. Mathematical convergence properties are analyzed for the filters using Lyapunov's stability theory. Lastly, we validate our proposed scheme by presenting simulation results. △ Less

Submitted 21 March, 2021; originally announced March 2021.

Comments: in IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2023

Music Genre Bars

Authors: Swaroop Panda, S. T. Roy

Abstract: Music Genres, as a popular meta-data of music, are very useful to organize, explore or search music datasets. Soft music genres are weighted multiple-genre annotations to songs. In this initial work, we propose horizontally stacked bar charts to represent a music dataset annotated by these soft music genres. For this purpose, we take an example of a toy dataset consisting of songs labelled with he… ▽ More Music Genres, as a popular meta-data of music, are very useful to organize, explore or search music datasets. Soft music genres are weighted multiple-genre annotations to songs. In this initial work, we propose horizontally stacked bar charts to represent a music dataset annotated by these soft music genres. For this purpose, we take an example of a toy dataset consisting of songs labelled with help of three music genres; Blues, Jazz and Country. We demonstrate how such a stacked bar chart can be used as a slider for user-input in an interface. We implement this by embedding this genre bar in a streaming application prototype and show its utility in choosing playlists. We finally conclude by proposing further work and future explorations on our proposed preliminary research. △ Less

Submitted 27 February, 2021; originally announced March 2021.

Comments: A version of this paper was presented at IEEE VIS 2020

Journal ref: IEEE VIS 2020 posters

Route Choice-based Socio-Technical Macroscopic Traffic Model

Authors: Tanushree Roy, Satadru Dey

Abstract: Human route choice is undeniably one of the key contributing factors towards traffic dynamics. However, most existing macroscopic traffic models are typically concerned with driving behavior and do not incorporate human route choice behavior models in their formulation. In this paper, we propose a socio-technical macroscopic traffic model that characterizes the traffic states using human route cho… ▽ More Human route choice is undeniably one of the key contributing factors towards traffic dynamics. However, most existing macroscopic traffic models are typically concerned with driving behavior and do not incorporate human route choice behavior models in their formulation. In this paper, we propose a socio-technical macroscopic traffic model that characterizes the traffic states using human route choice attributes. Essentially, such model provides a framework for capturing the Cyber-Physical-Social coupling in smart transportation systems. To derive this model, we first use Cumulative Prospect Theory (CPT) to model the human passengers' route choice under bounded rationality. These choices are assumed to be influenced by traffic alerts and other incomplete traffic information. Next, we assume that the vehicles are operating under a non-cooperative cruise control scenario. Accordingly, human route choice segregates the traffic into multiple classes where each class corresponds to a specific route between an origin-destination pair. Thereafter, we derive a Mean Field Game (MFG) limit of this non-cooperative game to obtain a macroscopic model which embeds the human route choice attribute. Finally, we analyze the mathematical characteristics of the proposed model and present simulation studies to illustrate the model behavior. △ Less

Submitted 26 January, 2021; originally announced January 2021.

Journal ref: in IEEE Transactions on Intelligent Transportation Systems, vol. 24, no. 6, pp. 5781-5791, June 2023

Anytime Planning: A Motion Planner for Dynamic Environment

Authors: Trishant Roy, Anindya Harchowdhury, Leena Vachhani

Abstract: Motion planning in the presence of multiple dynamic obstacles is an important research problem from the perspective of autonomous vehicles as well as space-constrained multi-robot work environment. In this paper, we address the motion planning problem for multiple dynamic obstacle rich environment and propose a probabilistically, complete novel motion planning algorithm. Our claim is that given a… ▽ More Motion planning in the presence of multiple dynamic obstacles is an important research problem from the perspective of autonomous vehicles as well as space-constrained multi-robot work environment. In this paper, we address the motion planning problem for multiple dynamic obstacle rich environment and propose a probabilistically, complete novel motion planning algorithm. Our claim is that given a fixed path cost i.e. the Euclidean path length, the proposed algorithm plans a path with the least computational time as compared to the state-of-the-art techniques. At the same time, given the time duration for planning, it plans the minimum cost path. Dynamic constraints have been taken into consideration while designing the planner such that the optimal planned path is feasible for implementation. The results of extensive simulation experiments show that the proposed sequential BIT* outperforms the DOVS both in the planned path length as well as the path generation time. △ Less

Submitted 26 December, 2019; originally announced December 2019.

Comments: 8 pages, 10 figures

Approximating the Void: Learning Stochastic Channel Models from Observation with Variational Generative Adversarial Networks

Authors: Timothy J. O'Shea, Tamoghna Roy, Nathan West

Abstract: Channel modeling is a critical topic when considering designing, learning, or evaluating the performance of any communications system. Most prior work in designing or learning new modulation schemes has focused on using highly simplified analytic channel models such as additive white Gaussian noise (AWGN), Rayleigh fading channels or similar. Recently, we proposed the usage of a generative adversa… ▽ More Channel modeling is a critical topic when considering designing, learning, or evaluating the performance of any communications system. Most prior work in designing or learning new modulation schemes has focused on using highly simplified analytic channel models such as additive white Gaussian noise (AWGN), Rayleigh fading channels or similar. Recently, we proposed the usage of a generative adversarial networks (GANs) to jointly approximate a wireless channel response model (e.g. from real black box measurements) and optimize for an efficient modulation scheme over it using machine learning. This approach worked to some degree, but was unable to produce accurate probability distribution functions (PDFs) representing the stochastic channel response. In this paper, we focus specifically on the problem of accurately learning a channel PDF using a variational GAN, introducing an architecture and loss function which can accurately capture stochastic behavior. We illustrate where our prior method failed and share results capturing the performance of such as system over a range of realistic channel distributions. △ Less

Submitted 20 August, 2018; v1 submitted 16 May, 2018; originally announced May 2018.

Precise Detection of Speech Endpoints Dynamically: A Wavelet Convolution based approach

Authors: Tanmoy Roy, Tshilidzi Marwala, Snehashish Chakraverty

Abstract: Precise detection of speech endpoints is an important factor which affects the performance of the systems where speech utterances need to be extracted from the speech signal such as Automatic Speech Recognition (ASR) system. Existing endpoint detection (EPD) methods mostly uses Short-Term Energy (STE), Zero-Crossing Rate (ZCR) based approaches and their variants. But STE and ZCR based EPD algorith… ▽ More Precise detection of speech endpoints is an important factor which affects the performance of the systems where speech utterances need to be extracted from the speech signal such as Automatic Speech Recognition (ASR) system. Existing endpoint detection (EPD) methods mostly uses Short-Term Energy (STE), Zero-Crossing Rate (ZCR) based approaches and their variants. But STE and ZCR based EPD algorithms often fail in the presence of Non-speech Sound Artifacts (NSAs) produced by the speakers. Algorithms based on pattern recognition and classification techniques are also proposed but require labeled data for training. A new algorithm termed as Wavelet Convolution based Speech Endpoint Detection (WCSEPD) is proposed in this article to extract speech endpoints. WCSEPD decomposes the speech signal into high-frequency and low-frequency components using wavelet convolution and computes entropy based thresholds for the two frequency components. The low-frequency thresholds are used to extract voiced speech segments, whereas the high-frequency thresholds are used to extract the unvoiced speech segments by filtering out the NSAs. WCSEPD does not require any labeled data for training and can automatically extract speech segments. Experiment results show that the proposed algorithm precisely extracts speech endpoints in the presence of NSAs. △ Less

Submitted 17 April, 2018; originally announced April 2018.

Comments: 25 Pages

Physical Layer Communications System Design Over-the-Air Using Adversarial Networks

Authors: Timothy J. O'Shea, Tamoghna Roy, Nathan West, Benjamin C. Hilburn

Abstract: This paper presents a novel method for synthesizing new physical layer modulation and coding schemes for communications systems using a learning-based approach which does not require an analytic model of the impairments in the channel. It extends prior work published on the channel autoencoder to consider the case where the channel response is not known or can not be easily modeled in a closed for… ▽ More This paper presents a novel method for synthesizing new physical layer modulation and coding schemes for communications systems using a learning-based approach which does not require an analytic model of the impairments in the channel. It extends prior work published on the channel autoencoder to consider the case where the channel response is not known or can not be easily modeled in a closed form analytic expression. By adopting an adversarial approach for channel response approximation and information encoding, we can jointly learn a good solution to both tasks over a wide range of channel environments. We describe the operation of the proposed adversarial system, share results for its training and validation over-the-air, and discuss implications and future work in the area. △ Less

Submitted 8 March, 2018; originally announced March 2018.

Over the Air Deep Learning Based Radio Signal Classification

Authors: Timothy J. O'Shea, Tamoghna Roy, T. Charles Clancy

Abstract: We conduct an in depth study on the performance of deep learning based radio signal classification for radio communications signals. We consider a rigorous baseline method using higher order moments and strong boosted gradient tree classification and compare performance between the two approaches across a range of configurations and channel impairments. We consider the effects of carrier frequency… ▽ More We conduct an in depth study on the performance of deep learning based radio signal classification for radio communications signals. We consider a rigorous baseline method using higher order moments and strong boosted gradient tree classification and compare performance between the two approaches across a range of configurations and channel impairments. We consider the effects of carrier frequency offset, symbol rate, and multi-path fading in simulation and conduct over-the-air measurement of radio classification performance in the lab using software radios and compare performance and training strategies for both. Finally we conclude with a discussion of remaining problems, and design considerations for using such techniques. △ Less

Submitted 12 December, 2017; originally announced December 2017.

Comments: 13 pages, 22 figures