Signal Processing II

B2L-B: Signal Processing II

Session Type: Lecture
Session Code: B2L-B
Location: Room 2
Date & Time: Thursday March 23, 2023 (10:20-11:20)
Chair: Tugba Erpek
Track: 3

Paper No.	Paper Name	Authors	Abstract
3046	Evaluating FFT Performance of the C and Rust Languages on Raspberry Pi Platforms	Michael P. Rooney Jr., Suzanne J. Matthews	The Fast Fourier Transform (FFT) is perhaps the most consequential algorithm for real-time applications for digital signals processing. Given the increased importance of securing devices on the edge, memory safety becomes an increasing concern for FFT applications. This work compares the performance of four FFT implementations written in the C and the Rust languages, benchmarked on the Raspberry Pi 4 and the Raspberry Pi Zero W platforms. Our results suggest that FFTs implemented in Rust are up to 45% more energy efficient than those written in C, and that Rust FFT implementations execute up to 37% faster than corresponding FFTs implemented in C. These results suggest that real-time application designers should take a closer look at the Rust language to enhance the safety and performance of their FFT applications.
3060	Toward Designing an Attentive Deep Trajectory Predictor Based on Bluetooth Low Energy Signal	Weijia Lu, Xiaofeng Ma, Xiaodong Zhang, Zhifei Yang, Qinghua Wang, Chuang Liu, Tao Yang	In this study, a novel attentive deep trajectory predictor is proposed for personal key (PK) localization problem in a Bluetooth low energy (BLE) network. This model has a unique sparseness design enlightened by the physical nature of the PK localization problem. Moreover, a set of geometrically inspired embedding losses are proposed to enhance model\'s generalization ability on different BLE anchor layout. Finally, the trained model with tiny footprint is deployed in a low-end vehicle processor. Intensive tests and carefully designed ablation studies are conducted to prove the robustness and effectiveness of the model.
3180	Prospect Theoretic Contract Design in a Stackelberg Game via Bayesian Estimation	Elham Jamalinia, Parv Venkitasubramaniam	Dynamic contract design with a cognitively biased principal is studied in this work using a Stackelberg game framework. Prospect theory is used to model the decision making of the cognitively biased principal. The goal of the principal is to design a sequence of contracts when faced with adverse selection-- lack of knowledge of agent\'s private information-- and moral hazard-- noisy observation of agent\'s effort. The principal observe the noisy output of the agent\'s effort and using the prior probabilities of the agent\'s private information, performs a Bayesian estimation of the agent\'s type taking into account Prospect theoretic probability weighting. The prospect theoretic decision making of principal is studied for a static game (one shot) and a dynamic game (finite number of rounds). It is demonstrated that the agent will benefit from principal\'s irrationality by extracting higher rent than the game with a rational principal. Theoretical results are derived for the binary type games, and the investigation is expanded to an $M$-ary hypothesis testing framework through numerical simulation results. The results demonstrate that in estimating the agent\'s type, a cognitively biased principal becomes more conservative to ensure agent\'s participation in the game than a rational principal thus allowing increased rent for the agent.