Fixed Sampling Rate Research Articles

Spectral network based on lattice convolution and adversarial training for noise-robust speech super-resolution.

Speech super-resolution aims to predict a high-resolution speech signal from its low-resolution counterpart. The previous models usually perform this task at a fixed sampling rate, reconstructing only high-frequency spectrogram components and merging them with low-frequency ones in noise-free cases. These methods achieve high accuracy, but they are less effective in real-world settings, where ambient noise and flexible sampling rates are presented. To develop a robust model that fits practical applications, in this work, we introduce Super Denoise Net (SDNet), a neural network for noise-robust super-resolution with flexible input sampling rates. To this end, SDNet's design includes gated and lattice convolution blocks for enhanced repair and temporal-spectral information capture. The frequency transform blocks are employed to model long frequency dependencies, and a multi-scale discriminator is proposed to facilitate the multi-adversarial loss training. The experiments show that SDNet outperforms current state-of-the-art noise-robust speech super-resolution models on multiple test sets, indicating its robustness and effectiveness in real-world scenarios.

Techniques for Detecting the Start and End Points of Sign Language Utterances to Enhance Recognition Performance in Mobile Environments

Recent AI-based technologies in mobile environments have enabled sign language recognition, allowing deaf individuals to communicate effectively with hearing individuals. However, varying computational performance across different mobile devices can result in differences in the number of image frames extracted in real time during sign language utterances. The number of extracted frames is a critical factor influencing the accuracy of sign language recognition models. If the number of extracted frames is too small, the performance of the sign language recognition model may decline. Additionally, detecting the start and end points of sign language utterances is crucial for improving recognition accuracy, as the period before the start point and after the end point often involves no action being performed. These parts do not capture the unique characteristics of each sign language. Therefore, this paper proposes a technique to dynamically adjust the sampling rate based on the number of frames extracted in real time during sign language utterances in mobile environments, with the aim of accurately detecting the start and end points of the sign language. Experiments were conducted to compare the proposed technique with the fixed sampling rate method and with the no-sampling method as a baseline. Our findings show that the proposed dynamic sampling rate adjustment method improves performance by up to 83.64% in top-5 accuracy and by up to 66.54% in top-1 accuracy compared to the fixed sampling rate method. The performance evaluation results underscore the effectiveness of our dynamic sampling rate adjustment approach in enhancing the accuracy and robustness of sign language recognition systems across different operational conditions.

Frequency-adaptive odd-harmonic repetitive control scheme for three-phase shunt active power filters

Repetitive control (RC), which can track the periodic input or suppress the periodic disturbance with a zero steady-state error, is a promising control strategy for shunt active power filters (SAPF). However, the conventional RC (CRC) scheme will suffer a severe performance degradation caused by the grid frequency variations. In this article, a frequency-adaptive odd-harmonic repetitive control scheme with a single and fixed sampling rate is proposed for SAPF to deal with grid frequency variations and provide fast transient response for RC system. The frequency-adaptability of proposed repetitive control scheme can be implemented by updating the coefficients of the approximate expression for the delay unit in repetitive controller. The FIR filter based on Lagrange linear interpolation is used to transform the multirate repetitive control system into single-rate repetitive control system to address the problems caused by the multirate repetitive control system. Moreover, a fractional-order linear compensator with a simple structure is designed to compensate the magnitude and phase of the system accurately. Compared with other classical repetitive control strategies, the proposed repetitive control scheme can provide a better adaptation to the steady-state deviation and dynamic variation of grid frequency and ensure the control performance of SAPF system. Simulation and experimental results verify the effectiveness of the proposed control scheme.

Convolutional neural network for improved event-based Shack-Hartmann wavefront reconstruction

Shack-Hartmann wavefront sensing is a technique for measuring wavefront aberrations, whose use in adaptive optics relies on fast position tracking of an array of spots. These sensors conventionally use frame-based cameras operating at a fixed sampling rate to report pixel intensities, even though only a fraction of the pixels have signal. Prior in-lab experiments have shown feasibility of event-based cameras for Shack-Hartmann wavefront sensing (SHWFS), asynchronously reporting the spot locations as log intensity changes at a microsecond time scale. In our work, we propose a convolutional neural network (CNN) called event-based wavefront network (EBWFNet) that achieves highly accurate estimation of the spot centroid position in real time. We developed a custom Shack-Hartmann wavefront sensing hardware with a common aperture for the synchronized frame- and event-based cameras so that spot centroid locations computed from the frame-based camera may be used to train/test the event-CNN-based centroid position estimation method in an unsupervised manner. Field testing with this hardware allows us to conclude that the proposed EBWFNet achieves sub-pixel accuracy in real-world scenarios with substantial improvement over the state-of-the-art event-based SHWFS. An ablation study reveals the impact of data processing, CNN components, and training cost function; and an unoptimized MATLAB implementation is shown to run faster than 800 Hz on a single GPU.

MotionID: Towards practical behavioral biometrics-based implicit user authentication on smartphones

Traditional one-time authentication mechanisms cannot authenticate smartphone users’ identities throughout the session — the concept of using behavioral-based biometrics captured by the built-in motion sensors and touch data is a candidate to solve this issue. Many studies proposed solutions for behavioral-based continuous authentication; however, they are still far from practicality and generality for real-world usage. To date, no commercially deployed implicit user authentication scheme exists because most of those solutions were designed to improve detection accuracy without addressing real-world deployment requirements. To bridge this gap, we tackle the limitations of existing schemes and reach towards developing a more practical implicit authentication scheme, dubbed MotionID, based on a one-class detector using behavioral data from motion sensors when users touch their smartphones. Compared with previous studies, our work addresses the following challenges: ① Global mobile average to dynamically adjust the sampling rate for sensors on any device and mitigate the impact of using sensors’ fixed sampling rate; ② Over-all-apps to authenticate a user across all the mobile applications, not only on-specific application; ③ Single-device-evaluation to measure the performance with multiple users’ (i.e., genuine users and imposters) data collected from the same device; ④ Rapid authentication to quickly identify users’ identities using a few samples collected within short durations of touching (1–5 s) the device; ⑤ Unconditional settings to collect sensor data from real-world smartphone usage rather than a laboratory study. To show the feasibility of MotionID for those challenges, we evaluated the performance of MotionID with ten users’ motion sensor data on five different smartphones under various settings. Our results show the impracticality of using a fixed sampling rate across devices that most previous studies have adopted. MotionID is able to authenticate users with an F1-score up to 98.5% for some devices under practical requirements and an F1-score up to roughly 90% when considering the drift concept and rapid authentication settings. Finally, we investigate time efficiency, power consumption, and memory usage considerations to examine the practicality of MotionID.

Weighted adaptive CUSUM mean chart with variable sample size and sampling interval

The classical CUSUM (C) chart is designed by assuming that the size of the shift in the process, in which a quick detection is important, is known a priori. However, this assumption is often violated in practice, as the actual shift size is rarely known even though it may vary within a certain interval. The adaptive C (AC) chart gives better protection against process shifts in a certain interval than the classical C chart. In this paper, we propose a new weighted AC (WAC) chart with the variable sample size and sampling interval (VSSI) feature, which encompasses the fixed sampling rate (FSR), variable sample size (VSS) and variable sampling interval (VSI) based WAC charts. The Monte Carlo simulations are adopted in the run-length computations of the considered charts, where the newly developed (VSS and VSSI based) charts outperform the existing (FSR and VSI) WAC charts when detecting small shifts in the process mean. An illustrative example shows the implementation of the existing and proposed WAC charts.

A Levenberg-Marquardt Algorithm for Sparse Identification of Dynamical Systems.

Low complexity of a system model is essential for its use in real-time applications. However, sparse identification methods commonly have stringent requirements that exclude them from being applied in an industrial setting. In this article, we introduce a flexible method for the sparse identification of dynamical systems described by ordinary differential equations. Our method relieves many of the requirements imposed by other methods that relate to the structure of the model and the dataset, such as fixed sampling rates, full state measurements, and linearity of the model. The Levenberg-Marquardt algorithm is used to solve the identification problem. We show that the Levenberg-Marquardt algorithm can be written in a form that enables parallel computing, which greatly diminishes the time required to solve the identification problem. An efficient backward elimination strategy is presented to construct a lean system model.

Wearables for running gait analysis: A study protocol.

Quantitative running gait analysis is an important tool that provides beneficial outcomes to injury risk/recovery or performance assessment. Wearable devices have allowed running gait to be evaluated in any environment (i.e., laboratory or real-world settings), yet there are a plethora of different grades of devices (i.e., research-grade, commercial, or novel multi-modal) available with little information to make informed decisions on selection. This paper outlines a protocol that will examine different grades of wearables for running gait analysis in healthy individuals. Specifically, this pilot study will: 1) examine analytical validity and reliability of wearables (research-grade, commercial, high-end multimodal) within a controlled laboratory setting; 2) examine analytical validation of different grades of wearables in a real-world setting, and 3) explore clinical validation and usability of wearables for running gait analysis (e.g., injury history (previously injured, never injured), performance level (novice, elite) and relationship to meaningful outcomes). The different grades of wearable include: (1) A research-grade device, the Ax6 consists of a configurable tri-axial accelerometer and tri-axial gyroscope with variable sampling capabilities; (2) attainable (low-grade) commercial with proprietary software, the DorsaVi ViMove2 consisting of two, non-configurable IMUs modules, with a fixed sampling rate and (3) novel multimodal high-end system, the DANU Sports System that is a pair of textile socks, that contain silicone based capacitive pressure sensors, and configurable IMU modules with variable sampling rates. Clinical trial registration: Trial registration: NCT05277181.

Power Gain from Energy Harvesting Sources at High MPPT Sampling Rates.

Energy harvesting (EH) sources require the tracking of their maximum power point (MPP) to ensure that maximum energy is captured. This tracking process, performed by an MPP tracker (MPPT), is performed by periodically measuring the EH transducer's output at a given sampling rate. The harvested power as a function of the sampling parameters has been analyzed in a few works, but the power gain achieved with respect to the case of a much slower sampling rate than the EH source's frequency has not been assessed so far. In this work, simple expressions are obtained that predict this gain assuming a Thévenin equivalent for the EH transducer. It is shown that the power gain depends on the relationship between the square of AC to DC open circuit voltage of the EH transducer. On the other hand, it is proven that harvested power increases, using a suitable constant signal for the MPP voltage instead of tracking the MPP at a low sampling rate. Experimental results confirmed the theoretical predictions. First, a function generator with a series resistor of 1 kΩ was used, emulating a generic Thévenin equivalent EH. Three waveform types were used (sinus, square, and triangular) with a DC voltage of 2.5 V and AC rms voltage of 0.83 V. A commercial MPPT with a fixed sampling rate of 3 Hz was used and the frequency of the waveforms was changed from 50 mHz to 50 Hz, thus effectively emulating different sampling rates. Experimental power gains of 11.1%, 20.7%, and 7.43% were, respectively, achieved for the sinus, square, and triangular waves, mainly agreeing with the theoretical predicted ones. Then, experimental tests were carried out with a wave energy converter (WEC) embedded into a drifter and attached to a linear shaker, with a sinus excitation frequency of 2 Hz and peak-to-peak amplitude of 0.4 g, in order to emulate the drifter's movement under a sea environment. The WEC provided a sinus-like waveform. In this case, another commercial MPPT with a sampling period of 16 s was used for generating a slow sampling rate, whereas a custom MPPT with a sampling rate of 60 Hz was used for generating a high sampling rate. A power gain around 20% was achieved in this case, also agreeing with the predicted gain.

Combination of Sequential Sampling Technique with GLR Control Charts for Monitoring Linear Profiles Based on the Random Explanatory Variables

Control charts play a beneficial role in the manufacturing process by reduction of non-compatible products and improving the final quality. In line with these aims, several adaptive methods in which samples can be taken with variable sampling rates and intervals have been proposed in the area of statistical process control (SPC). In some SPC applications, it is important to monitor a relationship between the response and independent variables—this is called profile monitoring. This article proposes adaptive generalized likelihood ratio (GLR) control charts based on variable sampling interval (VSI) and sequential sampling (SS) techniques for monitoring simple linear profiles. Because in some real-life problems, it may be possible that the user cannot control the values of explanatory variables; thus, in this paper, we focus on such a scenario. The performance of the proposed method is compared under three different situations, i.e., the fixed sampling rate (FSR), VSI, and SS, based on average time to signal (ATS) criteria for phase II analysis. Since the SS approach uses a novel sampling procedure based on the statistic magnitude, it has a superior performance over other competing charts. Several simulation studies indicate the superiority as the SS approach yields lower ATS values when there are single-step changes in the intercept, slope, standard deviation of the error term, and explanatory variables. In addition, some other related sensitivity analysis indicates that other aspects of the proposed methods, such as computational time, comparison with other control charts, and consideration of fixed explanatory variables. Furthermore, the results are supported by a real-life illustrative example from the adhesive manufacturing industry.

Driver impairment detection using decision tree based feature selection and classification

The driver's control authority plays a crucial role at commanding safely a vehicle that is the technical system. And the driver monitoring systems need to detect and identify whether the driver is impaired and able to percept and react possible hazards. In this paper, the driver's impairment detection system is presented. The procedure of fusing driving data constituted by a large set of vehicle sensors in addition to driver's eye glance location is elaborated. A batch data to train, validate and test the impairment detection system is constituted by co-existing 14 location information and 76 different sensors for trips that contain a crash or near crash instances. The procedure of aligning vehicle motion sensor data logged at fixed sampling rate with eye glance location information is presented by applying a sliding window approach. The window is slided when eye glance location is changed and mean of sensor values is used to present the batch data in this window. Each instance of the batch data is classified. A two stage detection algorithm is tested and evaluated. Each instance is classified as impair or non-impair using Extreme Gradient (xg) boosted trees and a voting mechanism for the instances is used to decide whether driver is impaired or not. The results illustrate the effectiveness of the impairment detection system.

A Multiple Interpolation Algorithm to Improve Resampling Accuracy in Data Triggers

To address the problem of low trigger accuracy during trigger resampling and variable sampling rate trigger resampling using a fixed sampling rate analog-to-digital converter (ADC), this paper proposes an interpolation method combining sinc interpolation and linear interpolation to improve accuracy, based on a digital trigger. After behavior simulation verification and actual field programmable gate array (FPGA) test verification, the data collected by two 3GSps 12-bit ADCs were subjected to 8-times sinc interpolation followed by 16-times linear interpolation processing, after which the original trigger resampling accuracy was increased by 128 times and the sampling rate could be realized to vary between 100 MHz and 1 GHz. A signal–noise ratio (SNR) of 46.80 dBFS, a spurious free dynamic range (SFDR) of 45.91 dB, and an effective number of bits (ENOB) of 7.48 bits were obtained by direct trigger resampling without algorithm processing in the behavior simulation. Meanwhile, an SNR of 58.98 dBFS, an SFDR of 60.96 dB, and an ENOB of 9.42 bits were obtained by trigger resampling after algorithm processing. Due to the influence of analog link signal loss and signal interference on the development board, an SNR, SFDR and ENOB of 51.97 dBFS, 61.26 dB, and 8.32 bits, respectively, were obtained from the trigger resampling in the FPGA test. The experimental results show that the algorithm has not only improved the triggering accuracy but has also improved the SNR, SFDR, and ENOB parameters.

Event- and sample-based audio processing: A shotgun marriage

Real-time audio computation tends to fall into two very different paradigms. On one hand we deal with continuous streams of audio sample frames having a fixed number of channels and a fixed data type. Our main conceptual tool is the Nyquist theorem. On the other hand, musical scores and sporadic control streams (for instance MIDI or network packets) do not fit comfortably into any fixed data type and need not fall on any fixed sample rate. In this talk I'll describe the choices taken in the design of the Pure Data real-time music and audio programming environment. The design makes it possible to make exactly reproducible real-time audio computations that combine two computational paradigms, one for vectorized data streams (intended for, but not limited to, audio), and the other to support control computations with sub-sample-accurate time tags.

Square wave reference digital lock-in detection using non-orthogonal demodulation

Digital lock-in detection technique is commonly used to measure the amplitude and phase of a selected frequency signal. The technique which uses a square wave as the reference signal has an advantage of easier implementation and higher computational efficiency compared to that uses a sine wave. However, one constraint for a square wave reference digital lock-in is that the sampling rate must be the integral multiple of 4 of all the signal frequencies. As the sampling clock may not always be able to set to the integral multiple of 4 of the signal frequencies, the constraint brings inconvenient for the implementation. Presented in this paper is a novel algorithm for square wave digital lock-in detection in which the sampling frequency doesn't have to meet the constraint. The algorithm allows frequency sweep measurements with a fixed sampling rate. For different relationships between the sampling rate and the signal frequency, different calculation equations are provided in this paper. Simulations and actual experiments show the feasibility of the proposed algorithm.

Dynamic Damping-Based Terminal Sliding Mode Event-Triggered Fault-Tolerant Pre-Compensation Stochastic Control for Tracked ROV

Due to the unknown disturbance caused by the harsh environment in deep water, the stability of Underwater Tracked Remotely Opreated Vehicle (UTROV) trajectory tracking control is affected; especially the resistance forces of random vibrations caused by non-differentiable random disturbance resistance, which has become one of the main problems in controller design. Considering engineering practice, a stochastic model and new dynamic damping-based terminal sliding mode event-triggered fault-tolerant controller were designed in this paper. Firstly, based on the random resistance pre-compensation theory for the first time, a stochastic model was designed for differential drive UTROV. Meanwhile, a new nonsingular terminal sliding mode and dynamic damping reaching law were designed to achieve global finite-time convergence and reduce chattering with better robust response speed. Furthermore, to deal with the wear and tear caused by actuator failure and fixed sampling rate transmission, a new dynamic event trigger mechanism was designed and the faults analyzed. On this basis, combined with the finite-time adaptive on-line estimation technology, it can not only better reduce the transmission frequency, but also the finite-time dynamic active fault-tolerant compensation. The control scheme has semi-globally finite-time stability in probability and is proved by theory, which is compliant with engineering requirements. Then, according to characteristics of innovation, the three groups of simulation of control methods are designed to compare the methods in this paper. Finally the advantages of the method are verified by simulation to achieve the design expectations.

A novel approach GRNTSTE to reconstruct gene regulatory interactions applied to a case study for rat pineal rhythm gene

Accurate inference and prediction of gene regulatory network are very important for understanding dynamic cellular processes. The large-scale time series genomics data are helpful to reveal the molecular dynamics and dynamic biological processes of complex biological systems. Firstly, we collected the time series data of the rat pineal gland tissue in the natural state according to a fixed sampling rate, and performed whole-genome sequencing. The large-scale time-series sequencing data set of rat pineal gland was constructed, which includes 480 time points, the time interval between adjacent time points is 3 min, and the sampling period is 24 h. Then, we proposed a new method of constructing gene expression regulatory network, named the gene regulatory network based on time series data and entropy transfer (GRNTSTE) method. The method is based on transfer entropy and large-scale time-series gene expression data to infer the causal regulatory relationship between genes in a data-driven mode. The comparative experiments prove that GRNTSTE has better performance than dynamical gene network inference with ensemble of trees (dynGENIE3) and SCRIBE, and has similar performance to TENET. Meanwhile, we proved that the performance of GRNTSTE is slightly lower than that of SINCERITIES method and better than other gene regulatory network construction methods in BEELINE framework, which is based on the BEELINE data set. Finally, the rat pineal rhythm gene expression regulatory network was constructed by us based on the GRNTSTE method, which provides an important reference for the study of the pineal rhythm mechanism, and is of great significance to the study of the pineal rhythm mechanism.

An adaptive multivariate EWMA mean chart with variable sample sizes and/or variable sampling intervals

AbstractNew adaptive multivariate EWMA (AME) charts based on the variable sample size (VSS), variable sampling interval (VSI) and variable sample size and sampling interval (VSSI) techniques are proposed to reduce the time of the existing AME chart in detecting a mean shift in the multivariate normal process. A Monte Carlo simulation is conducted to compare the time to signal characteristics of the new VSS/VSI/VSSI AME and existing fixed sampling rate (FSR) AME and VSSI multivariate EWMA (ME) charts. It is found that the VSS/VSI/VSSI AME charts generally surpass the existing FSR AME and the non‐adaptive VSSI ME charts in detecting small and moderate shifts in the mean of a multivariate normal process, in terms of the expected weighted time‐to‐signal and expected relative average time‐to‐signal performance measures. An example is given to show the implementation of the existing and proposed charts.

A parameter-free adaptive EWMA chart with variable sample sizes and variable sampling intervals for the process mean

In this article, a parameter-free adaptive EWMA (AE) chart, which adopts the variable sample size (VSS), variable sampling interval (VSI) or both (VSSVSI) features, is developed to monitor the mean of a normal process. The simulation method is employed to compute the run-length profiles of the proposed VSS, VSI and VSSVSI based AE charts. The extra quadratic loss and integral relative average run-length/time-to-signal measures are used as a performance criterion in order to evaluate the sensitivities of the control charts. The run-length evaluation shows that the proposed VSS, VSI and VSSVSI based AE charts surpass the VSS, VSI and VSSVSI based EWMA charts, respectively, in detecting small mean shifts but the latter prevails in detecting moderate to large mean shifts. The proposed charts also outperform their fixed sampling rate based AE counterparts. An example is given to compare the implementation of the existing and proposed charts.

Machine Learning with Variable Sampling Rate for Traffic Prediction in 6G MEC IoT

The high‐speed development of mobile broadband networks and IoT applications has brought about massive data transmission and data processing, and severe traffic congestion has adversely affected the fast‐growing networks and industries. To better allocate network resources and ensure the smooth operation of communications, predicting network traffic becomes an important tool. We investigate in detail the impact of variable sampling rate on traffic prediction and propose a high‐speed traffic prediction method using machine learning and recurrent neural networks. We first investigate a VSR‐NLMS adaptive prediction method to perform time series prediction dataset transformation. Then, we propose a VSR‐LSTM algorithm for real‐time prediction of network traffic. Finally, compared with the traditional traffic prediction algorithm based on fixed sampling rate (FSR‐LSTM), we simulate the prediction accuracy of the VSR‐LSTM algorithm based on the variable sampling rate proposed. The experiment shows that VSR‐LSTM has higher traffic prediction accuracy because its sampling rate varies with the traffic.

DANA

Motion sensors embedded in wearable and mobile devices allow for dynamic selection of sensor streams and sampling rates, enabling several applications, such as power management and data-sharing control. While deep neural networks (DNNs) achieve competitive accuracy in sensor data classification, DNN architectures generally process incoming data from a fixed set of sensors with a fixed sampling rate, and changes in the dimensions of their inputs cause considerable accuracy loss, unnecessary computations, or failure in operation. To address these limitations, we introduce a dimension-adaptive pooling (DAP) layer that makes DNNs flexible and more robust to changes in sensor availability and in sampling rate. DAP operates on convolutional filter maps of variable dimensions and produces an input of fixed dimensions suitable for feedforward and recurrent layers. Further, we propose a dimension-adaptive training (DAT) procedure for enabling DNNs that use DAP to better generalize over the set of feasible data dimensions at inference time. DAT comprises the random selection of dimensions during the forward passes and optimization with accumulated gradients of several backward passes. Combining DAP and DAT, we show how to transform existing non-adaptive DNNs into a Dimension-Adaptive Neural Architecture (DANA), while keeping the same number of parameters. Compared to existing approaches, our solution provides better average classification accuracy over the range of possible data dimensions at inference time and does not require up-sampling or imputation, thus reducing unnecessary computations. Experimental results on seven datasets (four benchmark real-world datasets for human activity recognition and three synthetic datasets) show that DANA prevents significant losses in classification accuracy of the state-of-the-art DNNs and, compared to baselines, it better captures correlated patterns in sensor data under dynamic sensor availability and varying sampling rates.