publications

2021

1. A wideband signal recognition dataset

   Nathan West, Timothy O’Shea, and Tamoghna Roy

   In 2021 IEEE 22nd International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), 2021

   Abs arXiv Bib

   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.

   @inproceedings{west2021wideband,
     title = {A wideband signal recognition dataset},
     author = {West, Nathan and O’Shea, Timothy and Roy, Tamoghna},
     booktitle = {2021 IEEE 22nd International Workshop on Signal Processing Advances in Wireless Communications (SPAWC)},
     pages = {6--10},
     year = {2021},
     organization = {IEEE},
   }

2. Wideband Signal Localization with Spectral Segmentation

   Nathan West, Tamoghna Roy, and Timothy O’Shea

   2021

   Abs arXiv Bib

   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.

   @article{west2021specsegment,
     author = {West, Nathan and Roy, Tamoghna and O'Shea, Timothy},
     keywords = {Signal Processing (eess.SP), FOS: Electrical engineering, electronic engineering, information engineering, FOS: Electrical engineering, electronic engineering, information engineering},
     title = {Wideband Signal Localization with Spectral Segmentation},
     publisher = {arXiv},
     year = {2021},
     copyright = {Creative Commons Attribution Non Commercial Share Alike 4.0 International},
     doi = {10.48550/ARXIV.2110.00583},
   }

3. A Semester Like No Other: Use of Natural Language Processing for Novice-Led Analysis on End-of-Semester Responses on Students’ Experience of Changing Learning Environments Due to COVID-19

   Sreyoshi Bhaduri, Michelle Soledad, Tamoghna Roy, and 2 more authors

   In 2021 ASEE Virtual Annual Conference Content Access, 2021

   Abs Bib PDF

   In response to campus closures due to COVID-19, the learning environment in a foundational engineering course unexpectedly shifted from hands-on, collaborative work to remote delivery, accomplished within a short period of time. Through end-of-semester course surveys, students were asked open-ended questions to get feedback about their experience with the goal of using student feedback for curriculum planning and improvement should there be continued need to facilitate the course remotely in subsequent semesters. However, with 1,170 responses, the volume of data made it challenging to analyze, interpret and use the feedback for decision-making for following semesters. To address this challenge, we utilized Natural Language Processing (NLP) based techniques - algorithmic ways to analyze, interpret, and present words and sentiments from student responses visually, to inform a novice-led analysis to ultimately help with course planning for future semesters.

   @inproceedings{bhaduri2021semester,
     title = {A Semester Like No Other: Use of Natural Language Processing for Novice-Led Analysis on End-of-Semester Responses on Students’ Experience of Changing Learning Environments Due to COVID-19},
     author = {Bhaduri, Sreyoshi and Soledad, Michelle and Roy, Tamoghna and Murzi, Homero and Knott, Tamara},
     booktitle = {2021 ASEE Virtual Annual Conference Content Access},
     year = {2021},
     doi = {10.18260/1-2--36609},
   }

2020

1. MSE Analysis of Bi-scale LMS Used for Narrowband Interference Cancellation

   Tamoghna Roy, Takeshi Ikuma, and AA Louis Beex

   In 2020 IEEE Latin-American Conference on Communications (LATINCOM), 2020

   Abs Bib PDF

   Adaptive LMS (Least Mean Square) equalizers are widely used in digital communication systems for their simplicity of implementation. Conventional adaptive filtering theory suggests that the upper bound on performance of such an equalizer is determined by the performance of a Wiener filter of the same structure. However, in the presence of a narrowband interferer the performance of the (normalized) LMS equalizer can be better than that of its Wiener counterpart. The Bi-scale NLMS (BLMS) algorithm enhances this NLMS (Normalized LMS) characteristic by simultaneously using two instantiations of NLMS that run at very different time scales. In this paper, the derivation of a predictive model for the MSE (Mean Square Error) performance of the BLMS equalizer as narrowband interference canceler is shown. The predictive model can be used to adjust canceler parameters on the fly without the delay needed for time-consuming simulations. Simulation results validate the proposed MSE model, which is shown to predict performance of the BLMS equalizer over a wide range of parameters.

   @inproceedings{roy2020mse,
     title = {MSE Analysis of Bi-scale LMS Used for Narrowband Interference Cancellation},
     author = {Roy, Tamoghna and Ikuma, Takeshi and Beex, AA Louis},
     booktitle = {2020 IEEE Latin-American Conference on Communications (LATINCOM)},
     pages = {1--6},
     year = {2020},
     organization = {IEEE},
     doi = {10.1109/LATINCOM50620.2020.9282346},
   }

2019

1. Generative adversarial radio spectrum networks

   Tamoghna Roy, Tim O’Shea, and Nathan West

   In Proceedings of the ACM Workshop on Wireless Security and Machine Learning, 2019

   Abs Bib PDF

   Simulating and imitating RF communications signals and systems is a core function of jammers, spoofers, and other attacks in wireless radio environments which seek to confuse spectrum users as to what is occurring in the spectrum around them. Replay attacks and "DRFMs" have long been commonly used to deceive and probe radio systems, however generative models introduce an interesting new angle wherein generative replay can now produce examples of signals of similar structure and properties to arbitrary signals which are not verbatim replays and which may be varied in an infinite number of ways. Further, as GANs have demonstrated a strong ability to learn distributions from complex scenes and datasets, we consider the task of full-band spectral generation in addition to single signal generation to validate and demonstrate the feasibility of such an approach, to refine the algorithmic approach, and to quantify and illustrate the capabilities of such an approach on modern day signal sets.

   @inproceedings{roy2019generative,
     title = {Generative adversarial radio spectrum networks},
     author = {Roy, Tamoghna and O'Shea, Tim and West, Nathan},
     booktitle = {Proceedings of the ACM Workshop on Wireless Security and Machine Learning},
     pages = {12--15},
     year = {2019},
     doi = {10.1145/3324921.3328782},
   }

2. Approximating the void: Learning stochastic channel models from observation with variational generative adversarial networks

   Timothy J O’Shea, Tamoghna Roy, and Nathan West

   In 2019 International Conference on Computing, Networking and Communications (ICNC), 2019

   Abs arXiv Bib

   Channel modeling is a critical topic when considering accurately designing or evaluating the performance of a communications system. Most prior work in designing or learning new modulation schemes has focused on using simplified analytic channel models such as additive white Gaussian noise (AWGN), Rayleigh fading channels or other similar compact parametric models. In this paper, we extend recent work training generative adversarial networks (GANs) to approximate wireless channel responses to more accurately reflect the probability distribution functions (PDFs) of stochastic channel behaviors. We introduce the use of variational GANs to provide appropriate architecture and loss functions which accurately capture these stochastic behaviors. Finally, we illustrate why prior GAN-based methods failed to accurately capture these behaviors and share results illustrating the performance of such as system over a range of complex realistic channel effects.

   @inproceedings{o2019approximating,
     title = {Approximating the void: Learning stochastic channel models from observation with variational generative adversarial networks},
     author = {O’Shea, Timothy J and Roy, Tamoghna and West, Nathan},
     booktitle = {2019 International Conference on Computing, Networking and Communications (ICNC)},
     pages = {681--686},
     year = {2019},
     organization = {IEEE},
     doi = {10.1109/ICCNC.2019.8685573},
   }

2018

1. SigMF: The signal metadata format

   Ben Hilburn, Nathan West, Tim O’Shea, and 1 more author

   In Proceedings of the GNU Radio Conference, 2018

   Abs Bib PDF

   The Signal Metadata Format (SigMF) specifies a way to describe sets of recorded digital signal samples with metadata written in JSON. SigMF can be used to describe general information about a collection of samples, the characteristics of the system that generated the samples, and features of the signal itself. It is designed to be a simple and portable format that is easily used by memory-limited applications with real-time requirements and minimal dependencies.

   @inproceedings{hilburn2018sigmf,
     title = {SigMF: The signal metadata format},
     author = {Hilburn, Ben and West, Nathan and O'Shea, Tim and Roy, Tamoghna},
     booktitle = {Proceedings of the GNU Radio Conference},
     volume = {3},
     number = {1},
     year = {2018},
   }

2. Demonstrating deep learning based communications systems over the air in practice

   Timothy J O’Shea, Tamoghna Roy, Nathan West, and 1 more author

   In 2018 IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN), 2018

   Abs Bib PDF

   Several novel approaches to wireless communications system design have recently been introduced which use deep learning to synthesize and adapt a new class of signal processing systems to the actual data and effects present in the radio environment. Both the autoencoder-based communications system and the feature learning-based radio signal sensor represent significant progress in the ability of radio systems to optimize directly on real-world data samples and distributions. As part of this demonstration, we will show two real-world systems operating over-the-air using these approaches, which we are rapidly maturing at DeepSig to bring theory into practice.

   @inproceedings{o2018demonstrating,
     title = {Demonstrating deep learning based communications systems over the air in practice},
     author = {O’Shea, Timothy J and Roy, Tamoghna and West, Nathan and Hilburn, Benjamin C},
     booktitle = {2018 IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN)},
     pages = {1--2},
     year = {2018},
     organization = {IEEE},
     doi = {10.1109/DySPAN.2018.8610491},
   }

3. Power Measurement Based Code Classification for Programmable Logic Circuits

   Tamoghna Roy, and AA Louis Beex

   In 2018 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT), 2018

   Abs Bib PDF

   Traditional cyber security and monitoring systems rely on prior knowledge about possible attacks, which renders them ineffective against novel schemes (zero-day attacks). Moreover, intruders are targeting different types of devices (e.g. Programmable Logic Circuits/Controllers (PLC)) for which traditional security systems are unavailable. A solution to this problem is provided by analysis of the power consumption behavior of a system and relating power consumption characteristics to behavior internal to the device. In this work, a detection system is developed which is capable of discriminating between different codes executed by the target device, a PLC, based on passive measurement at the external power supply point; this is unlike previous methods, which relied on a specific sensor located directly over the chip. Results show that for a PLC, which is executing four different codes, the classification system can discriminate with very high accuracy (> 99%).

   @inproceedings{roy2018power,
     title = {Power Measurement Based Code Classification for Programmable Logic Circuits},
     author = {Roy, Tamoghna and Beex, AA Louis},
     booktitle = {2018 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT)},
     pages = {644--648},
     year = {2018},
     organization = {IEEE},
     doi = {10.1109/ISSPIT.2018.8642680},
   }

4. Over-the-air deep learning based radio signal classification

   Timothy James O’Shea, Tamoghna Roy, and T Charles Clancy

   IEEE Journal of Selected Topics in Signal Processing, 2018

   Abs arXiv Bib

   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 multipath fading in simulation, and conduct over-the-air measurement of radio classification performance in the lab using software radios, and we compare performance and training strategies for both. Finally, we conclude with a discussion of remaining problems, and design considerations for using such techniques.

   @article{o2018over,
     title = {Over-the-air deep learning based radio signal classification},
     author = {O’Shea, Timothy James and Roy, Tamoghna and Clancy, T Charles},
     journal = {IEEE Journal of Selected Topics in Signal Processing},
     volume = {12},
     number = {1},
     pages = {168--179},
     year = {2018},
     publisher = {IEEE},
     doi = {10.1109/JSTSP.2018.2797022},
   }

5. Physical layer communications system design over-the-air using adversarial networks

   Timothy J O’Shea, Tamoghna Roy, Nathan West, and 1 more author

   In 2018 26th European Signal Processing Conference (EUSIPCO), 2018

   Abs arXiv Bib

   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 autoencoderto consider the case where the stochastic channel response is not known or can not be easily modeled in a closed form analytic expression. By adopting an adversarial approach for learning a channel response approximation and information encoding, we jointly learn a solution to both tasks applicable 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.

   @inproceedings{o2018physical,
     title = {Physical layer communications system design over-the-air using adversarial networks},
     author = {O'Shea, Timothy J and Roy, Tamoghna and West, Nathan and Hilburn, Benjamin C},
     booktitle = {2018 26th European Signal Processing Conference (EUSIPCO)},
     pages = {529--532},
     year = {2018},
     organization = {IEEE},
     doi = {10.23919/EUSIPCO.2018.8553233},
   }

6. Non-Wiener Characteristics of LMS Adaptive Equalizers: A Bit Error Rate Perspective

   Tamoghna Roy

   Virginia Tech, 2018

   Bib PDF

   @phdthesis{roy2018non,
     title = {Non-Wiener Characteristics of LMS Adaptive Equalizers: A Bit Error Rate Perspective},
     author = {Roy, Tamoghna},
     year = {2018},
     school = {Virginia Tech},
   }

2017

1. Learning robust general radio signal detection using computer vision methods

   Tim O’Shea, Tamoghna Roy, and T Charles Clancy

   In 2017 51st asilomar conference on signals, systems, and computers, 2017

   Abs Bib PDF

   We introduce a new method for radio signal detection and localization within the time-frequency spectrum based on the use of convolutional neural networks for bounding box regression. Recently, this class of approach has surpassed human-level performance on computer vision benchmarks for object detection, but similar techniques have not yet been adopted for radio applications. We introduce the basic approach explain how labeled training data containing wideband spectrum annotated with masks and bounding boxes can be used to train a highly effective radio signal detector which achieves higher levels of contextual understanding and improved sensitivity performance when compared with more traditional nave energy thresholding based signal detection schemes. We extend prior work from the computer vision field, employing a variation of the You Only Look Once (YOLO) architecture which is a fast and accurate visual object detector. Results are shown from illustrating the effectiveness from our entry into the DARPA Battle-of-the-ModRecs competition and over the air datasets

   @inproceedings{o2017learning,
     title = {Learning robust general radio signal detection using computer vision methods},
     author = {O'Shea, Tim and Roy, Tamoghna and Clancy, T Charles},
     booktitle = {2017 51st asilomar conference on signals, systems, and computers},
     pages = {829--832},
     year = {2017},
     organization = {IEEE},
     doi = {10.1109/ACSSC.2017.8335463},
   }

2. Spectral detection and localization of radio events with learned convolutional neural features

   Timothy J O’Shea, Tamoghna Roy, and Tugba Erpek

   In 2017 25th European Signal Processing Conference (EUSIPCO), 2017

   Abs Bib PDF

   We introduce a method for detecting, localizing and identifying radio transmissions within wide-band time-frequency power spectrograms using feature learning using convolutional neural networks on their 2D image representation. By doing so we build a foundation for higher level contextual radio spectrum event understanding, labeling, and reasoning in complex shared spectrum and many-user environments by developing tools which can rapidly understand and label sequences of events based on experience and labeled data rather than signal-specific detection algorithms such as matched filters.

   @inproceedings{o2017spectral,
     title = {Spectral detection and localization of radio events with learned convolutional neural features},
     author = {O'Shea, Timothy J and Roy, Tamoghna and Erpek, Tugba},
     booktitle = {2017 25th European Signal Processing Conference (EUSIPCO)},
     pages = {331--335},
     year = {2017},
     organization = {IEEE},
     doi = {10.23919/EUSIPCO.2017.8081223},
   }

3. A word-space visualization approach to study college of engineering mission statements

   Sreyoshi Bhaduri, and Tamoghna Roy

   In 2017 IEEE Frontiers in Education Conference (FIE), 2017

   Abs Bib PDF

   Most higher education institutions have a mission statement that is developed strategically by the institutions and often reflect the college’s unique mission which sets it apart from peer institutions. Through this study, we describe the use of a Machine Learning and Natural Language Processing based textual data analytics to understand the word choices in the mission statements of U.S. based colleges of engineering. Our purpose is to understand the key similarities and differences between the choice of words used in the mission statements of the two groups: public colleges and private colleges of engineering. We were specifically interested in studying the terms related to diversity and inclusion and see the trends in the use of specific terms relating to diverse communities, intersections, minority populations, and the like. In this research study, we used a Word2Vec approach to visualize the words from mission statements for 59 colleges of engineering in the United States. The contribution of this research is in the form of a visualization mapping the vector space model for word usage and complete vocabulary of pertinent words from the statements analyzed. The preliminary results of this study will help inform current state of vocabulary used in mission statements in the colleges of engineering across the United States. Ultimately, such analyses can help administrators in the development of strategies on the formation of mission and vision statements for universities by allowing insight into vocabulary currently used, to understand what words/terms may not be adequately addressed. Additionally, we are excited to present a contemporary textual data analytical technique of Natural Language Processing using a word vector representation tool such as word2vec for analyzing textual data in the field of engineering education.

   @inproceedings{bhaduri2017word,
     title = {A word-space visualization approach to study college of engineering mission statements},
     author = {Bhaduri, Sreyoshi and Roy, Tamoghna},
     booktitle = {2017 IEEE Frontiers in Education Conference (FIE)},
     pages = {1--5},
     year = {2017},
     organization = {IEEE},
     doi = {10.1109/FIE.2017.8190704},
   }

4. Demonstrating use of natural language processing to compare college of engineering mission statements

   Sreyoshi Bhaduri, and Tamoghna Roy

   In 2017 ASEE Annual Conference & Exposition, 2017

   Abs Bib PDF

   Most higher education institutions have a mission and/or vision statement that is designed to communicate with a variety of audiences. These statements are developed strategically by organizations and often reflect the college’s unique vision which sets it apart from peer institutions. Analytical techniques which rely on word usage, semantic information, and metadata information can be used to generate powerful descriptive models with allow us to obtain relevant information from text-based data. This work-in-progress study presents a Natural Language Processing based textual data analytical approach to study the mission and vision statements with the purpose of understanding the key similarities and differences between the choice of words used in them. We analyzed a total of 59 engineering colleges: 29 public, and 30 private, across the United States. Results of this study indicate that there is indeed a difference in word frequencies for public versus private engineering colleges. The contribution of this research is in the form of charts quantitatively summarizing the comparative word usage and a descriptive overview of the complete vocabulary of pertinent words from the statements analyzed. Topical clustering based on words seen in prior literature was also conducted to analyze comparative categories across the institutions. This study can help inform strategies on the formation of mission and vision statements for universities by allowing administrators insight into vocabulary used across colleges.

   @inproceedings{bhaduri2017demonstrating,
     title = {Demonstrating use of natural language processing to compare college of engineering mission statements},
     author = {Bhaduri, Sreyoshi and Roy, Tamoghna},
     booktitle = {2017 ASEE Annual Conference \& Exposition},
     year = {2017},
     doi = {10.18260/1-2--28102},
   }

2015

1. Non-Cooperative Localization Using Differential RSS and Link Loss Parameter Estimation

   Tamoghna Roy, and A. A. (Louis) Beex

   In 2015 Wireless Innovation Forum Conference on Communications Technologies and Software Defined Radio (WInnComm 2015), 2015

   Bib

   @inproceedings{roy2015non,
     title = {Non-Cooperative Localization Using Differential RSS and Link Loss Parameter Estimation},
     author = {Roy, Tamoghna and (Louis) Beex, A. A.},
     booktitle = {2015 Wireless Innovation Forum Conference on Communications Technologies and Software Defined Radio (WInnComm 2015)},
     pages = {118--123},
     year = {2015},
   }

2014

1. BER Modeling for Interference Canceling Adaptive NLMS Equalizer

   Tamoghna Roy

   Virginia Tech, 2014

   Bib PDF

   @phdthesis{roy2014ber,
     title = {BER Modeling for Interference Canceling Adaptive NLMS Equalizer},
     author = {Roy, Tamoghna},
     year = {2014},
     school = {Virginia Tech},
   }

2. Non Cooperative Localization Under Spatially Correlated Shadowing Using Weighted Differential RSS

   Tamoghna Roy, and A.A. (Louis) Beex

   In 2014 Wireless Innovation Forum Conference on Communications Technologies and Software Defined Radio (SDR-WInnComm 2014), 2014

   Bib

   @inproceedings{roy2014non,
     title = {Non Cooperative Localization Under Spatially Correlated Shadowing Using Weighted Differential RSS},
     author = {Roy, Tamoghna and (Louis) Beex, A.A.},
     booktitle = {2014 Wireless Innovation Forum Conference on Communications Technologies and Software Defined Radio (SDR-WInnComm 2014)},
     pages = {228--233},
     year = {2014},
   }

3. LOCATION VERIFICATION USING DIFFERENTIAL RECEIVED SIGNAL STRENGTH

   Tamoghna Roy, AA Louis Beex, and Jeffrey H Reed

   In FCC SAS Workshop, 2014

   Bib

   @inproceedings{roy2014location,
     title = {LOCATION VERIFICATION USING DIFFERENTIAL RECEIVED SIGNAL STRENGTH},
     author = {Roy, Tamoghna and Beex, AA Louis and Reed, Jeffrey H},
     booktitle = {FCC SAS Workshop},
     year = {2014},
   }

2013

1. BER modeling for interference canceling FIR Wiener equalizer

   Tamoghna Roy, and A.A. (Louis) Beex

   In International Conference on Computing, Networking and Communications (ICNC), 2013, 2013

   Abs Bib PDF

   The performance of a narrowband interference canceling FIR Wiener equalizer is analyzed. While mean squared error (MSE) relates to bit error rate (BER), their connection is not necessarily a direct one when the detector output noise is not Gaussian. We show that BER can be increasing for increasing signal power (or decreasing noise power) even though MSE is decreasing. A Gaussian BER model may not be accurate then. For digital modulation schemes using FIR Wiener equalizers in a narrowband interference dominated environment, a Gaussian sum model is derived for the Wiener filter output. The analytical evaluation of the probability of bit error based on the Gaussian sum model produces a BER prediction that is shown to provide a close match with observed/estimated BER, in particular for lower order equalizers.

   @inproceedings{roy2013ber,
     title = {BER modeling for interference canceling FIR Wiener equalizer},
     author = {Roy, Tamoghna and (Louis) Beex, A.A.},
     booktitle = {International Conference on Computing, Networking and Communications (ICNC), 2013},
     pages = {278--282},
     year = {2013},
     doi = {10.1109/ICCNC.2013.6504095},
   }