Lingfeng Xu

Mr. Lingfeng Xu


Research Focus & Metrics

Academic Career Length

16Years (2011-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
25 September 2018 in Sensors
Verified
Journal Article
Open access
PDF file
LXLingfeng Xu
Lingfeng Xu
XCXiang Chen
Xiang Chen
SCShuai Cao
Shuai Cao
XZXu Zhang
Xu Zhang
XCXun Chen
Xun Chen
To find out the feasibility of different neural networks in sEMG-based force estimation, in this paper, three types of networks, namely convolutional neural network (CNN), long short-term memory (LSTM) network and their combination (C-LSTM) were applied to predict muscle force generated in static isometric elbow flexion across three different circumstances (multi-subject, subject-dependent and subject-independent). Eight healthy men were recruited for the experiments, and the results demonstrated that all the three models were applicable for force estimation, and LSTM and C-LSTM achieved better performances. Even under subject-independent situation, they maintained mean RMSE% of as low as 9.07 ± 1.29 and 8.67 ± 1.14. CNN turned out to be a worse choice, yielding a mean RMSE% of 12.13 ± 1.98. To our knowledge, this work was the first to employ CNN, LSTM and C-LSTM in sEMG-based force estimation, and the results not only prove the strength of the proposed networks, but also pointed out a potential way of achieving high accuracy in real-time, subject-independent force estimation.

1 January 2023 in JASA Express Letters
Verified
Journal Article
Open access
PDF file
LXLingfeng Xu
Lingfeng Xu
JLJulie M Liss
Julie M Liss
VBVisar Berisha
Visar Berisha
Studies have shown deep neural networks (DNN) as a potential tool for classifying dysarthric speakers and controls. However, representations used to train DNNs are largely not clinically interpretable, which limits clinical value. Here, a model with a bottleneck layer is trained to jointly learn a classification label and four clinically-interpretable features. Evaluation of two dysarthria subtypes shows that the proposed method can flexibly trade-off between improved classification accuracy and discovery of clinically-interpretable deficit patterns. The analysis using Shapley additive explanation shows the model learns a representation consistent with the disturbances that define the two dysarthria subtypes considered in this work.

1 January 2016 in Acta Physica Sinica
Verified
Journal Article
Closed access
PDF file
LXLingfeng Xu
Lingfeng Xu
CLChuandong Li
Chuandong Li
CLChen Ling
Chen Ling
In recent years,the modeling and application of biological neurons have gained more and more attention.By now, the research on neuron models has become one of the most important branches of neuroscience.Neuron models can be used in various areas,such as biomimetic applications,memory design,logical computing,and signal processing. Furthermore,it is significant to study the dynamic characteristics of neural system by using neuron models.In this paper,the historical development of neuron models is reviewed.The neuron models have experienced three development stages.In the pioneering stage,a group of scientists laid the experimental and theoretical foundation for later research. Then,the whole study started to blossom after the publication of Hodgkin-Huxley model.In the 1970s and 1980s,various models were proposed.One of the research focuses was the simulation of neural repetitive spiking.Since the 1990s, researchers have paid more attention to setting up models that are both physiologically meaningful and computationally effective.The model put forward by Izhikevich E M has been proved to solve the problem successfully.Recently,IBM presented a versatile spiking neuron model based on 1272 ASIC gates.The model,both theoretically understandable and physically implementable,has been used as a basic building block in IBM's neuro-chip TrueNorth.In the paper, seventeen neuron models worth studying are listed.To give a more explicit explanation,these models are classified as two groups,namely conductance-dependent and conductance-independent models.The former group's goal is to model the electrophysiology of neuronal membrane,while the latter group is only to seek for capturing the input-output behavior of a neuron by using simple mathematical abstractions.The complexity and features of each model are illustrated in a chart,while the prominent repetitive spiking curves of each model are also exhibited.Five of the models are further detailed,which are the Hodgkin-Huxley model,the Integrate-and-fire model,the Fitzhugh-Nagumo model,the Izhikevich model,and the most recent model used by IBM in its neuro-chip TrueNorth.Finally,three questions are put forward at the end of the paper,which are the most important problems that today's researchers must consider when setting up new neuron models.In conclusion,the feasibility of physical implementation remains to be a challenge to all researchers. Through the aforementioned work,the authors aim to provide a reference for the study that follows,helping researchers to compare those models in order to choose the properest one.

1 October 2018 in IEEE Transactions on Neural Systems and Rehabilitation Engineering
Verified
Journal Article
Closed access
LXLingfeng Xu
Lingfeng Xu
XCXiang Chen
Xiang Chen
SCShuai Cao
Shuai Cao
XZXu Zhang
Xu Zhang
XCXun Chen
Xun Chen
To alleviate the negative impacts of muscle fatigue on force estimation model, a modification framework taking use of fatigue index was put forward in this paper. Muscle force and surface electromyography (sEMG) were first collected using high-density electrode grid and dynamometer. Then, multi-step signal pre-processing as well as a nonnegative matrix factorization (NMF) based signal optimization were conducted, with fatigue indices being extracted in the same time. Next, a degree 4 polynomial fitting model was employed to undertake the training process, and the relationship between the generated model parameters and fatigue indices was built up. In the end, the parameter-index relationship was applied on different testing sets to complete fatigue-modified force estimation. Significant improvement was found in most testing cases across different sexes and ages. Relative decreases of 36.5%, 20.7% and 20.4% in the percentage root mean square error (RMSE%) were achieved by young males, young females and elderly males. The proposed method can boost the performances of force estimation models, thereby contributing to the development of a variety of fields including biomechanical study, rehabilitation treatment, and prosthesis research.

5 December 2024 in Journal of Alzheimer’s Disease
Verified
Journal Article
Closed access
LXLingfeng Xu
Lingfeng Xu
KCKewei Chen
Kewei Chen
KMKimberly D Mueller
Kimberly D Mueller
JLJulie Liss
Julie Liss
VBVisar Berisha
Visar Berisha
Background: Mild cognitive impairment (MCI) has been recognized as a possible precursor to Alzheimer’s disease (AD). Recent research focusing on connected speech has uncovered various features strongly correlated with MCI due to AD and related dementias. Despite these advancements, the impact of early cognitive decline on articulatory precision has not been thoroughly investigated. Objective: This study introduced the phoneme log-likelihood ratio (PLLR) to assess the articulatory precision of speakers across different cognitive status levels and compared its effectiveness to existing well-studied acoustic features. Methods: The study utilized speech recordings from a picture description task, which included data from 324 cognitively unimpaired participants with low amyloid-β burden (CU, Aβ( − )), 47 cognitively unimpaired participants with high amyloid-β burden (CU, Aβ( + )), 69 participants with MCI, and 20 participants with dementia. Nine acoustic features were extracted from the speech recordings, covering three categories: speech fluency, speech pace, and articulatory precision. Welch’s t -test and Hedge’s g were adopted to assess their discriminative ability. Results: A reduction in speech fluency and pace was observed among participants in the MCI and dementia groups. The PLLR shows large effect sizes in distinguishing between cognitively unimpaired participants with low Aβ burden and those diagnosed with MCI or dementia. Additionally, the test-retest reliability experiment indicated moderate repeatability of the features under study. Conclusions: The study reveals PLLR as a sensitive indicator capable of detecting subtle articulatory variations across groups, while also providing further support for the association between reduced articulatory precision and early cognitive decline.

1 November 2015
Verified
Conference Paper
Closed access
LXLingfeng Xu
Lingfeng Xu
CLChuandong Li
Chuandong Li
LCLing Chen
Ling Chen
Since its discovery, memristor has been well studied by researchers from all around the world, and its application in recognition proves to be very promising. In this paper, we modify a memristor crossbar circuit from an existing work to recognize 8 × 8 pixel binary images. We use analog memristors instead of binary memristors to complete the circuit. The simulated recognition rate is 82.5% in average, and we step further by carrying out a Monte Carlo simulation to analyze the performances of the circuit under different memristance variations and statistical distributions. We find that as the memristance variation rises up, the recognition rate under Gaussian distribution drops quickly, while the performance under uniform distribution is relatively stable. In the final part, we provide some outlooks and remarks on the possible improvements of the circuit.