Search

This paper analyzes the impact of imperfect communication channels on decentralized federated learning (D-FL) and subsequently determines the optimal number of local aggregations per training round, adapting to the network topology and imperfect channels. We start by deriving the bias of locally aggregated D-FL models under imperfect channels from the ideal global models requiring perfect channels and aggregations. The bias reveals that excessive local aggregations can accumulate communication errors and degrade convergence. Another important aspect is that we analyze a convergence upper bound of D-FL based on the bias. By minimizing the bound, the optimal number of local aggregations is identified to balance a trade-off with accumulation of communication errors in the absence of knowledge of the channels. With this knowledge, the impact of communication errors can be alleviated, allowing the convergence upper bound to decrease throughout aggregations. Experiments validate our convergence analysis and also identify the optimal number of local aggregations on two widely considered image classification tasks. It is seen that D-FL, with an optimal number of local aggregations, can outperform its potential alternatives by over 10% in training accuracy.

This document contains the appendices for our paper titled ``Performance Bounds for Passive Sensing in Asynchronous ISAC Systems." The appendices include rigorous derivations of key formulas, detailed proofs of the theorems and propositions introduced in the paper, and details of the algorithm tested in the numerical simulation for validation. These appendices aim to support and elaborate on the findings and methodologies presented in the main text. All external references to equations, theorems, and so forth, are directed towards the corresponding elements within the main paper., Comment: 5 pages

Recently, there has been growing interest in investigating the effects of self-representation on user experience and perception in virtual environments. However, few studies investigated the effects of levels of body representation (full-body, lower-body and viewpoint) on locomotion experience in terms of spatial awareness, self-presence and spatial presence during virtual locomotion. Understanding such effects is essential for building new virtual locomotion systems with better locomotion experience. In the present study, we first built a walking-in-place (WIP) virtual locomotion system that can represent users using avatars at three levels (full-body, lower-body and viewpoint) and is capable of rendering walking animations during in-place walking of a user. We then conducted a virtual locomotion experiment using three levels of representation to investigate the effects of body representation on spatial awareness, self-presence and spatial presence during virtual locomotion. Experimental results showed that the full-body representation provided better virtual locomotion experience in these three factors compared to that of the lower-body representation and the viewpoint representation. The lower-body representation also provided better experience than the viewpoint representation. These results suggest that self-representation of users in virtual environments using a full-body avatar is critical for providing better locomotion experience. Using full-body avatars for self-representation of users should be considered when building new virtual locomotion systems and applications., Comment: 2023 IEEE International Conference on Systems, Man, and Cybernetics (SMC), Hyatt Maui, Hawaii, USA, Oct 1-4, 2023

Body actions and head gestures are natural interfaces for interaction in virtual environments. Existing methods for in-place body action recognition often require hardware more than a head-mounted display (HMD), making body action interfaces difficult to be introduced to ordinary virtual reality (VR) users as they usually only possess an HMD. In addition, there lacks a unified solution to recognize in-place body actions and head gestures. This potentially hinders the exploration of the use of in-place body actions and head gestures for novel interaction experiences in virtual environments. We present a unified two-stream 1-D convolutional neural network (CNN) for recognition of body actions when a user performs walking-in-place (WIP) and for recognition of head gestures when a user stands still wearing only an HMD. Compared to previous approaches, our method does not require specialized hardware and/or additional tracking devices other than an HMD and can recognize a significantly larger number of body actions and head gestures than other existing methods. In total, ten in-place body actions and eight head gestures can be recognized with the proposed method, which makes this method a readily available body action interface (head gestures included) for interaction with virtual environments. We demonstrate one utility of the interface through a virtual locomotion task. Results show that the present body action interface is reliable in detecting body actions for the VR locomotion task but is physically demanding compared to a touch controller interface. The present body action interface is promising for new VR experiences and applications, especially for VR fitness applications where workouts are intended., Comment: 2023 IEEE International Conference on Virtual Reality and 3D User Interfaces, Shanghai, Mar 25-29, 2023

Classification is an important step in machine vision systems; it reveals the true identity of an object using features extracted in pre-processing steps. Practical usage requires the operation to be fast, energy efficient and easy to implement. In this paper, we present a design of the Minimum Distance Classifier based on an FPGA platform. It is optimized by the pipelined structure to strike a balance between device utilization and computational speed. In addition, the dimension of the feature space is modeled as a generic parameter, making it possible for the design to re-generate hardware to cope with feature space with arbitrary dimensions. Its primary application is demonstrated in color segmentation on FPGA in the form of efficient classification using color as a feature. This result is further extended by introducing a multi-class component labeling module to label the segmented color components and measure their geometric properties. The combination of these two modules can effectively detect road signs as the region of interests., Comment: Updated version

Generating forward-backward self-representation leg animation in virtual environments for walking-in-place (WIP) techniques is an underexplored research topic. A challenging aspect of the problem is to find an appropriate mapping from tracked vertical foot motion to natural cyclical movements of real walking. In this work, we present a kinematic approach based on animation rigging to generating real-time leg animation. Our method works by tracking vertical in-place foot movements of a user with a Kinect v2 sensor and mapping tracked foot height to inverse kinematics (IK) targets. These IK targets were aligned with an avatar's feet to guide the virtual feet to perform cyclic walking motions. We conducted a user study to evaluate our approach. Results showed that the proposed method produced compelling forward-backward leg animation during walking. We show that the proposed technique can be easily integrated into existing WIP techniques., Comment: Accepted by ACM SIGGRAPH VRCAI 2022

Hand gesture is a new and promising interface for locomotion in virtual environments. While several previous studies have proposed different hand gestures for virtual locomotion, little is known about their differences in terms of performance and user preference in virtual locomotion tasks. In the present paper, we presented three different hand gesture interfaces and their algorithms for locomotion, which are called the Finger Distance gesture, the Finger Number gesture and the Finger Tapping gesture. These gestures were inspired by previous studies of gesture-based locomotion interfaces and are typical gestures that people are familiar with in their daily lives. Implementing these hand gesture interfaces in the present study enabled us to systematically compare the differences between these gestures. In addition, to compare the usability of these gestures to locomotion interfaces using gamepads, we also designed and implemented a gamepad interface based on the Xbox One controller. We conducted empirical studies to compare these four interfaces through two virtual locomotion tasks. A desktop setup was used instead of sharing a head-mounted display among participants due to the concern of the Covid-19 situation. Through these tasks, we assessed the performance and user preference of these interfaces on speed control and waypoints navigation. Results showed that user preference and performance of the Finger Distance gesture were close to that of the gamepad interface. The Finger Number gesture also had close performance and user preference to that of the Finger Distance gesture. Our study demonstrates that the Finger Distance gesture and the Finger Number gesture are very promising interfaces for virtual locomotion. We also discuss that the Finger Tapping gesture needs further improvements before it can be used for virtual walking.

Gesture awareness plays a crucial role in promoting human-computer interface. Previous works either depend on customized hardware or need a priori learning of wireless signal patterns, facing downsides in terms of the privacy concern, availability and reliability. In this paper, we propose CentiTrack, the first centimeter-level passive gesture tracking system that works with only three commodityWiFi devices, without any extra hardware modifications or wearable sensors. To this end, we first identify the Channel State Information (CSI) measurement error sources in the physical layer process, and then denoise CSI by the complex ratio between adjacent antennas. Principal Component Analysis (PCA) is further adopted to separate the reflected signals from noises. Benchmark experiments are conducted to verify that the phase changes of denoised CSI are proportional to the length changes of dynamic path reflected off the hand. In addition, we adopt the Multiple Signal Classification (MUSIC) algorithm to estimate the Angle-of-Arrivals (AoAs) of dynamic paths, and then locate the initial position of hands with triangulation. We also propose a novel static componnets elimination algorithm for tracking correction by eliminating the components unrelated to motion. A prototype of CentiTrack is fully realized and evaluated in various real scenarios. Extensive experiments show that CentiTrack is superior in terms of tracking accuracy, sensing range and device cost, compared with the state-of-the-arts.