19 February 2016 in IEEE Transactions on Signal and Information Processing over Networks
This paper presents a graph signal denoising method with the trilateral filter defined in the graph spectral domain. The original trilateral filter (TF) is a data-dependent filter that is widely used as an edge-preserving smoothing method for image processing. However, because of the data-dependency, one cannot provide its frequency domain representation. To overcome this problem, we establish the graph spectral domain representation of the data-dependent filter, i.e., a spectral graph TF (SGTF). This representation enables us to design an effective graph signal denoising filter with a Tikhonov regularization. Moreover, for the proposed graph denoising filter, we provide a parameter optimization technique to search for a regularization parameter that approximately minimizes the mean squared error w.r.t. the unknown graph signal of interest. Comprehensive experimental results validate our graph signal processing-based approach for images and graph signals.
29 August 2017 in IEEE Transactions on Signal Processing
We propose an approximation method for thresholding of singular values using Chebyshev polynomial approximation (CPA). Many signal processing problems require iterative application of singular value decomposition (SVD) for minimizing the rank of a given data matrix with other cost functions and/or constraints, which is called matrix rank minimization. In matrix rank minimization, singular values of a matrix are shrunk by hard-thresholding, soft-thresholding, or weighted soft-thresholding. However, the computational cost of SVD is generally too expensive to handle high-dimensional signals, such as images; hence, in this case, matrix rank minimization requires enormous computation time. In this paper, we leverage CPA to (approximately) manipulate singular values without computing singular values and vectors. The thresholding of singular values is expressed by a multiplication of certain matrices, which is derived from a characteristic of CPA. The multiplication is also efficiently computed using the sparsity of signals. As a result, the computational cost is significantly reduced. Experimental results suggest the effectiveness of our method through several image processing applications based on matrix rank minimization with nuclear norm relaxation in terms of computation time and approximation precision.
11 April 2023 in JMIR Research Protocols
Background: Digital health technologies using mobile apps and wearable devices are a promising approach to the investigation of substance use in the real world and for the analysis of predictive factors or harms from substance use. Moreover, consecutive repeated data collection enables the development of predictive algorithms for substance use by machine learning methods. Objective: We developed a new self-monitoring mobile app to record daily substance use, triggers, and cravings. Additionally, a wearable activity tracker (Fitbit) was used to collect objective biological and behavioral data before, during, and after substance use. This study aims to describe a model using machine learning methods to determine substance use. Methods: This study is an ongoing observational study using a Fitbit and a self-monitoring app. Participants of this study were people with health risks due to alcohol or methamphetamine use. They were required to record their daily substance use and related factors on the self-monitoring app and to always wear a Fitbit for 8 weeks, which collected the following data: (1) heart rate per minute, (2) sleep duration per day, (3) sleep stages per day, (4) the number of steps per day, and (5) the amount of physical activity per day. Fitbit data will first be visualized for data analysis to confirm typical Fitbit data patterns for individual users. Next, machine learning and statistical analysis methods will be performed to create a detection model for substance use based on the combined Fitbit and self-monitoring data. The model will be tested based on 5-fold cross-validation, and further preprocessing and machine learning methods will be conducted based on the preliminary results. The usability and feasibility of this approach will also be evaluated. Results: Enrollment for the trial began in September 2020, and the data collection finished in April 2021. In total, 13 people with methamphetamine use disorder and 36 with alcohol problems participated in this study. The severity of methamphetamine or alcohol use disorder assessed by the Drug Abuse Screening Test-10 or the Alcohol Use Disorders Identification Test-10 was moderate to severe. The anticipated results of this study include understanding the physiological and behavioral data before, during, and after alcohol or methamphetamine use and identifying individual patterns of behavior. Conclusions: Real-time data on daily life among people with substance use problems were collected in this study. This new approach to data collection might be helpful because of its high confidentiality and convenience. The findings of this study will provide data to support the development of interventions to reduce alcohol and methamphetamine use and associated negative consequences. International Registered Report Identifier (IRRID): DERR1-10.2196/44275