Linear Discriminant Classifier Research Articles

Enhancing Speaker Recognition Models with Noise-Resilient Feature Optimization Strategies

This paper delves into an in-depth exploration of speaker recognition methodologies, with a primary focus on three pivotal approaches: feature-level fusion, dimension reduction employing principal component analysis (PCA) and independent component analysis (ICA), and feature optimization through a genetic algorithm (GA) and the marine predator algorithm (MPA). This study conducts comprehensive experiments across diverse speech datasets characterized by varying noise levels and speaker counts. Impressively, the research yields exceptional results across different datasets and classifiers. For instance, on the TIMIT babble noise dataset (120 speakers), feature fusion achieves a remarkable speaker identification accuracy of 92.7%, while various feature optimization techniques combined with K nearest neighbor (KNN) and linear discriminant (LD) classifiers result in a speaker verification equal error rate (SV EER) of 0.7%. Notably, this study achieves a speaker identification accuracy of 93.5% and SV EER of 0.13% on the TIMIT babble noise dataset (630 speakers) using a KNN classifier with feature optimization. On the TIMIT white noise dataset (120 and 630 speakers), speaker identification accuracies of 93.3% and 83.5%, along with SV EER values of 0.58% and 0.13%, respectively, were attained utilizing PCA dimension reduction and feature optimization techniques (PCA-MPA) with KNN classifiers. Furthermore, on the voxceleb1 dataset, PCA-MPA feature optimization with KNN classifiers achieves a speaker identification accuracy of 95.2% and an SV EER of 1.8%. These findings underscore the significant enhancement in computational speed and speaker recognition performance facilitated by feature optimization strategies.

Analyzing lower body movements using machine learning to classify autistic children

Autism spectrum disorder is a neurological disorder that affects children at an early age and its symptoms appear in varying degrees. Early detection of ASD and follow-up of treatment by specialists help improve autistic people's lives and reduce the symptoms they suffer from. In the past few years, a group of studies conducted gait analysis to diagnose children with autism, and although the results of these studies showed weak motor skills of Autism Spectrum Disorder children (ASD) compared to their peers of Typically Developing children (TD), there are some challenges facing researchers in this field. The first is the ability to create A safe, effective, friendly environment that reduces tension in ASD children and helps classify children into ASD/TD effectively. And the second is the ability to create an automated ASD diagnosis system. In this work, a system based on machine learning and augmented reality technology has been proposed to create an effective and friendly interactive environment that helps diagnose ASD and classify children to ASD / TD. The movement of the lower limbs of 15 ASD children and 15 TD children are tracked via a marker-less technique through a Kinect v2 sensor when playing a specific game based on kicking the ball task, that is using augmented reality technology. Then the kinematics features are extracted through Linear Discriminant Analysis technology. To determine these features' ability in the classification process, three machine learning algorithms are used, namely Support Vector Machine, Linear Discriminant Analysis, and Decision Tree. The results showed that the obtained kinematics features can classify children into ASD/TD, where the Support Vector Machine classifier obtained a high accuracy of 96.7% with sensitivity equal to 93.3% and specificity equal to 100%. In addition, the Linear Discriminant Analysis classifier obtained less accuracy of 93.3% with sensitivity equal to 93.3% and specificity equal to 93.3% compared with the Support Vector Machine classifier. Where the Decision Tree classifier obtained the lowest accuracy of 80.0% with sensitivity equal to 73.3% and specificity equal to 86.7% compared with the Support Vector Machine and Linear Discriminant Analysis classifiers. In addition to that, when using Linear Discriminant Analysis to reduce dimensions and extract features, the accuracy, sensitivity, and specificity of the Support Vector Machine classifier were increased and achieved by 100%. Based on these results, it has been proven that the proposed method is effective in classifying children to ASD/TD.

Abstract 2582: Intra- and peri-tumoral radiomic features are predictive of pathologic response to multiple neoadjuvant therapy regimen in rectal cancers via pre-treatment MRI

Abstract Introduction: Rectal cancers undergo neoadjuvant chemoradiation prior to resection, with increasing evidence that neoadjuvant chemotherapy or immunotherapy can boost clinical or pathologic response rates. However, baseline clinical tumor staging and carcinoembryonic antigen levels are limited for identifying which patients will most benefit from a multimodal neoadjuvant regimen, and which will not. We evaluated whether radiomics (computational features from radiographic images) from pretreatment MRI could identify which rectal cancer patients will achieve clinical or pathologic response to different neoadjuvant treatment regimens. Methods: MRI scans were retrospectively obtained from 3 different institutions for rectal cancer patients who had either undergone neoadjuvant chemoradiation or an experimental immunotherapy + chemoradiation. After pre-processing all MRI scans, tumor was manually delineated with an automated annotation of a 10 mm peritumor region. 916 radiomic features were extracted from two planes of acquisition (axial, coronal) and two ROIs (tumor, peritumor). 5 top-ranked radiomic features were identified from each region and plane via multi-stage mutual information feature selection to train a linear discriminant machine classifier to assign a predicted likelihood of achieving complete response to each patient. Model performance was validated via ROC analysis against response groups defined via (i) tumor regression grade indicating no tumor present on surgical pathology, and (ii) 1-year clinical complete response for patients who underwent non-operative management. Clinical variables (CEA, clinical stage) were statistically compared between response groups. Results: Training cohort comprised 64 patients who underwent chemoradiation alone (2 institutions, median age 61 yrs, M:F:MtF = 36:27:1). Validation cohort included 37 patients who underwent experimental TGFß inhibitor therapy prior to chemotherapy + chemoradiation (1 institution, median age 51 yrs, M:F = 25:12). Area under the ROC curve for multi-plane tumor+peritumor radiomic features (0.765 ± 0.054) was significantly higher for distinguishing complete response vs non/partial response compared to tumor (AUC = 0.635 ± 0.057, p < 0.0001) or peritumor (AUC = 0.612 ± 0.064, p < 0.0001). In holdout validation, performance was maintained with AUCs of 0.742 (tumor) 0.596 (peritumor), and 0.700 (all), respectively. Baseline CEA levels (p=0.3385) and cT stage (p=0.3386) were found to lack statistical significance as a predictor of response. Conclusion: Radiomic features from tumor and peritumor regions on pre-treatment MRI scans in rectal cancer patients may enable robust and accurate identification of pathologic or near-complete response after multiple neoadjuvant treatment regimen, while outperforming baseline clinical variables. Citation Format: Leo Bao, Thomas DeSilvio, Benjamin N. Parker, Mohsen Hariri, Prathyush Chirra, Murad Labbad, Stephen Tang, Gregory M. O'Connor, Emily Steinhagen, Jennifer L. Miller-Ocuin, Amit Gupta, Eric L. Marderstein, Aaron Carroll, Marka Crittenden, Michael J. Gough, Smitha Krishnamurthi, Kristina H. Young, Satish E. Viswanath. Intra- and peri-tumoral radiomic features are predictive of pathologic response to multiple neoadjuvant therapy regimen in rectal cancers via pre-treatment MRI [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2582.

Drought susceptibility mapping in Iraq using GRACE/GRACE-FO, GLDAS, and machine learning algorithms

Drought susceptibility mapping is crucial for mitigating water scarcity and its cascading effects. This study utilizes machine learning (ML) and Gravity Recovery and Climate Experiment (GRACE) data to identify drought-prone areas in Iraq. Terrestrial water storage anomaly (TWSA) data derived from GRACE observations was utilized, in conjunction with various meteorological and topographical factors (temperature, precipitation, wind speed, potential evapotranspiration, aridity index, normalized difference vegetation index, groundwater storage, root zone moisture, elevation, and slope) sourced from diverse global datasets, primarily those comprising the Global Land Data Assimilation System (GLDAS). These combined datasets served as input for five machine learning models to identify the drought prone area of Iraq: linear discriminant analysis, classification and regression trees, support vector machine, random forest, and k-nearest neighbor. All models exhibited exceptional performance (>0.9 accuracy, >0.8 kappa values, >0.9 areas under the receiver operating characteristic curve) highlighting the effectiveness of the combined TWSA and multi-factor approach. The analysis identified five drought susceptibility zones across Iraq, ranging from very low to very high. An area of approximately 160,650 km2, primarily concentrated in the northeastern and Mesopotamian plains of central and southern Iraq, was identified as exhibiting high to very high drought susceptibility, highlighting the necessity for targeted mitigation strategies. This study pioneers the use of GRACE-derived TWSA and GLDAS data for drought susceptibility mapping in Iraq, paving the way for data-driven water management interventions.

Feature extraction and classification of different hand movements from the emg signal using linear discriminant analysis classifier

In biomedical research, Electromyography (EMG) data play a crucial role as a bridge between human motions and machine interpretation, offering valuable insights into muscle activation. EMG signals give vital information on hand movements in the context of applications like gesture recognition, prosthetic control, and rehabilitation. This paper describes the classification of EMG signals based on muscle motions, which makes it simpler to identify distinct gestures or movements. A Linear Discriminant Analysis (LDA) classifier is used to differentiate between various classes of muscle activity. In order to record EMG signals during hand motions, surface electrodes are carefully positioned on pertinent muscles. Muscle activity may be tracked in real time with these non-invasive electrodes. In order to extract meaningful information from these signals, which are complex and frequently contaminated by noise, strong feature extraction techniques are needed. When working with noisy signals, denoising is a commonly used approach to restoring the original quality of the source data. It attempts to maintain relevant information by reducing noise in the raw EMG signals. In order to retrieve only the pertinent information from the original EMG signal data, any unnecessary noise must first be removed. Through the identification of key characteristics in the time, frequency, and time-frequency domains, it transforms unstructured EMG data. This procedure improves the next step of classification, which is the identification and classification of patterns in the EMG signals. Ultimately, the obtained information is employed to classify signals by the Linear Discriminant Analysis (LDA) classifier, demonstrating a distinction between various muscle motions with over 80% accuracy.

Low Dimensional Multi Class Steganalysis of Spatial LSB based Stego Images Using Textural Features

Image steganalysis ranges from detecting the presence of covert information in an image (passive steganalysis) to extraction of the information from the stego image (active steganalysis). One of the steps in active steganalysis is determining the stego algorithm used to produce the stego image. In this paper, a low dimensional combination of textural features is adapted for steganalysis. Also a novel blind statistical steganalyser to determine the spatial domain Least Significant Bit (LSB) based algorithms using one against one multi class classification is proposed. The proposed steganalyser is a multiclass ensemble Fisher Linear Discriminant (FLD) classifier that uses novel low dimensional textural features for steganalysis. The performed experiments on the Bossbase database for 5 different LSB based algorithms for 8 different payloads show that the results are much better than the state of art steganalyser.

Test

Image steganalysis ranges from detecting the presence of covert information in an image (passive steganalysis) to extraction of the information from the stego image (active steganalysis). One of the steps in active steganalysis is determining the stego algorithm used to produce the stego image. In this paper, a low dimensional combination of textural features is adapted for steganalysis. Also a novel blind statistical steganalyser to determine the spatial domain Least Significant Bit (LSB) based algorithms using one against one multi class classification is proposed. The proposed steganalyser is a multi class ensemble Fisher Linear Discriminant (FLD) classifier that uses novel low dimensional textural features for steganalysis. The performed experiments on the Bossbase database for 5 different LSB based algorithms for 8 different payloads show that the results are much better than the state of art steganalyser

Harvest Forecasting Improvement Using Federated Learning and Ensemble Model

Harvest forecasting is the great demand of multiple aspects like temperature, rain, environment, and their relations. The existing study investigates the climate conditions and aids the cultivators to know the harvest yields before planting in farms. The proposed study uses federated learning. In addition, the additional widespread techniques such as bagging classifier, extra tees classifier, linear discriminant analysis classifier, quadratic discriminant analysis classifier, stochastic gradient boosting classifier, blending models, random forest regressor, and AdaBoost are utilized together. These presented nine algorithms achieved exemplary satisfactory accuracies. The powerful contributions of proposed algorithms can create exact harvest forecasting. Ultimately, we intend to compare our study with the earlier research's results.

Detection of attacks on software defined networks using machine learning techniques and imbalanced data handling methods

AbstractSoftware‐defined networks (SDNs) have gained popularity in recent years as a solution for the fundamental issues that affect traditional dispersed networks. The primary advantage of SDNs is the decoupling of the control plane from the data plane, which increases the flexibility of the network. The SDN represents a network architecture of the next generation, however, its configuration options are centralized, leaving it open for cyber‐attacks. This paper concentrates on the early identification of attacks in an SDN environment. When malicious traffic is affecting in an SDN topology, an artificial intelligence (AI) module in the topology is used to detect the attack and stop the attack source using machine learning (ML) techniques. The architecture presented in this research allows for the comparison of several ML classification techniques that are used to identify different sorts of network attacks. For attack detection, eight ML techniques are used, namely logistic regression (LR), linear discriminant analysis (LDA), Naïve Bayes (NB), k‐nearest neighbor (KNN), classification and regression tree (CART), AdaBoost (AB), random forest (RF), and support‐vector machine (SVM) classifiers. These techniques are tested on the InSDN dataset, which is a novel attack‐specific SDN dataset. The results show that the highest accuracy of 98.6% is achieved with the LDA classifier. Further improvement in the accuracy of classification models is observed when random over‐sampling, synthetic minority oversampling technique (SMOTE), random under‐sampling, and under‐sampling with Tomek links and near‐miss concept are applied to address the class imbalance problem. After applying these methods, the LDA classifier showed an accuracy of 98.79%.

Machine learning for biomedical-focus on brain signals

There is a growing interest in machine learning (ML) in this decade. This growing interest is accelerated by cheaper computing power and low-cost memory. Thus, large amount of data can be stored, processed and analyzed efficiently. Machine learning has used in brain machine systems (BMS) system that converts brain impulses into messages or commands. In this paper we propose an EEG-based BMS with the focus on evoked potential. An average classification accuracy of 95% was attained among nine participants. With a rate of 4 flashes per second implemented, selecting one of four possibilities takes 5 s, resulting in an information transfer rate of 24 bits/min. Also, brain computer interfacing using oscillatory activity was measured. The results show that after around 5 h of co-adaptive training over many days, the average 3-class accuracy of the Linear Discriminant Analysis committee classifier reached about 80%, with a false positive rate for motor imagery recognition of around 17%.

The influence of food values on snacks and beverage choices: an experimental study in Mexico

This study examines whether Mexican consumers consider the value associated with the attributes of manufactured snacks when deciding what to eat between meals. A choice experiment involving the selection of healthy versus unhealthy snacks and beverages was performed using Qualtrics. A set of four options, each one corresponding to a combination of mid-day snack and beverage, was presented to 256 Mexican consumers, who, after making a choice, specified the importance they assigned to food choice values. Linear discriminant analysis and classification trees were used to identify what food choice values and psychographic and demographic traits explain the choices. The analytical results indicate weight control, healthiness and sustainability, and the emotional value of healthy snacks mainly drives the choices of health-conscious consumers.

Localization model and rank-based features selection approach for the classification of GGO and consolidation stages of COVID-19

Diseases like Middle East respiratory syndrome (MERS) and severe acute respiratory syndrome (SARS) have been caused by a large family virus known as Coronavirus. High-resolution computed tomography (HRCT) chest is used as a useful tool for the triage and stratification of COVID-19-suspected patients having respiratory symptoms until they are confirmed by polymerase chain reaction (PCR) test. To further expedite the reporting of HRCT chest in a short time, computer-aided detection algorithms can be more useful. In this work, a method is proposed that consists of three phases including COVID-19 classification, localization, and segmentation. The COVID-19-ResNet-50 model is designed for COVID-19 classification using HRCT slices. The features are obtained from a pre-trained ResNet-50 model, and the process of identifying the most pertinent features is carried out using the Wilcoxon Separability Criteria (WSC). The linear discriminant (LD) classifier is used to classify COVID-19 providing an accuracy of 0.99. To localize the COVID-19-infected region, the classified HRCT slices are fed as input to an open neural network exchange (ONNX) that is a backbone of the Tiny-You Look at Once (YOLO)-v2 model. The proposed COVID-19-YOLOv2 model localized the COVID-19 lesions with 0.948 intersections over union (IoU). A three-dimensional semantic neural network comprising 10 selected layers is proposed whose training based on the best-selected hyperparameters is performed, it accurately segments the localized infected regions having an 0.98 IoU. The proposed method is accessed on national and international datasets that provide improved performance as compared to the most recent works published in this domain.

Enhancement of Classifier Performance Using Swarm Intelligence in Detection of Diabetes from Pancreatic Microarray Gene Data.

In this study, we focused on using microarray gene data from pancreatic sources to detect diabetes mellitus. Dimensionality reduction (DR) techniques were used to reduce the dimensionally high microarray gene data. DR methods like the Bessel function, Discrete Cosine Transform (DCT), Least Squares Linear Regression (LSLR), and Artificial Algae Algorithm (AAA) are used. Subsequently, we applied meta-heuristic algorithms like the Dragonfly Optimization Algorithm (DOA) and Elephant Herding Optimization Algorithm (EHO) for feature selection. Classifiers such as Nonlinear Regression (NLR), Linear Regression (LR), Gaussian Mixture Model (GMM), Expectation Maximum (EM), Bayesian Linear Discriminant Classifier (BLDC), Logistic Regression (LoR), Softmax Discriminant Classifier (SDC), and Support Vector Machine (SVM) with three types of kernels, Linear, Polynomial, and Radial Basis Function (RBF), were utilized to detect diabetes. The classifier's performance was analyzed based on parameters like accuracy, F1 score, MCC, error rate, FM metric, and Kappa. Without feature selection, the SVM (RBF) classifier achieved a high accuracy of 90% using the AAA DR methods. The SVM (RBF) classifier using the AAA DR method for EHO feature selection outperformed the other classifiers with an accuracy of 95.714%. This improvement in the accuracy of the classifier's performance emphasizes the role of feature selection methods.

Lateralization and Time-Course of Cortical Phonological Representations during Syllable Production.

Spoken language contains information at a broad range of timescales, from phonetic distinctions on the order of milliseconds to semantic contexts which shift over seconds to minutes. It is not well understood how the brain's speech production systems combine features at these timescales into a coherent vocal output. We investigated the spatial and temporal representations in cerebral cortex of three phonological units with different durations: consonants, vowels, and syllables. Electrocorticography (ECoG) recordings were obtained from five participants while speaking single syllables. We developed a novel clustering and Kalman filter-based trend analysis procedure to sort electrodes into temporal response profiles. A linear discriminant classifier was used to determine how strongly each electrode's response encoded phonological features. We found distinct time-courses of encoding phonological units depending on their duration: consonants were represented more during speech preparation, vowels were represented evenly throughout trials, and syllables during production. Locations of strongly speech-encoding electrodes (the top 30% of electrodes) likewise depended on phonological element duration, with consonant-encoding electrodes left-lateralized, vowel-encoding hemispherically balanced, and syllable-encoding right-lateralized. The lateralization of speech-encoding electrodes depended on onset time, with electrodes active before or after speech production favoring left hemisphere and those active during speech favoring the right. Single-electrode speech classification revealed cortical areas with preferential encoding of particular phonemic elements, including consonant encoding in the left precentral and postcentral gyri and syllable encoding in the right middle frontal gyrus. Our findings support neurolinguistic theories of left hemisphere specialization for processing short-timescale linguistic units and right hemisphere processing of longer-duration units.

Autonomic nervous system modulation during self-induced non-ordinary states of consciousness

Self-induced cognitive trance (SICT) is a voluntary non-ordinary state of consciousness characterized by a lucid yet narrowed awareness of the external surroundings. It involves a hyper-focused immersive experience of flow, expanded inner imagery, modified somatosensory processing, and an altered perception of self and time. SICT is gaining attention due to its potential clinical applications. Similar states of non-ordinary state of consciousness, such as meditation, hypnosis, and psychedelic experiences, have been reported to induce changes in the autonomic nervous system. However, the functioning of the autonomic nervous system during SICT remains poorly understood. In this study, we aimed to investigate the impact of SICT on the cardiac and respiratory signals of 25 participants proficient in SICT. To accomplish this, we measured various metrics of heart rate variability (HRV) and respiration rate variability (RRV) in three conditions: resting state, SICT, and a mental imagery task. Subsequently, we employed a machine learning framework utilizing a linear discriminant analysis classifier and a cross-validation scheme to identify the features that exhibited the best discrimination between these three conditions. The results revealed that during SICT, participants experienced an increased heart rate and a decreased level of high-frequency (HF) HRV compared to the control conditions. Additionally, specific increases in respiratory amplitude, phase ratio, and RRV were observed during SICT in comparison to the other conditions. These findings suggest that SICT is associated with a reduction in parasympathetic activity, indicative of a hyperarousal state of the autonomic nervous system during SICT.

Raman spectroscopic diagnosis of blast-induced traumatic brain injury in rats combined with machine learning

Blast-induced traumatic brain injury (bTBI) is a kind of nervous system disease, which results in a major health and economic problem to society. However, the rapid and label-free detection method with high sensitivity is still in great demand for the diagnosis of bTBI, especially for mild bTBI. In this paper, we report a new strategy for bTBI diagnosis through hippocampus and hypothalamus tissues based on Raman spectroscopy. The spectral characteristics of hippocampus and hypothalamus tissues of experimental bTBI in rats have been investigated for mild and moderate degrees at 3 h, 6 h, 24 h, 48 h, 72 h after blast exposure. The results show that the Raman spectra of mild and moderate bTBIs in 300–1700 cm−1 and 2800–3000 cm−1 regions exhibit significant differences at different time points compared with the control group. The main reason is the content change of proteins and lipids in hippocampus and hypothalamus tissues after bTBI. Moreover, four machine learning algorithms are used to automatically identify mild and moderate bTBIs at different time points (a total of 11 groups). The highest diagnostic accuracies are up to 95.3% and 88.5% based on Raman spectra of hippocampus and hypothalamus tissue, respectively. In addition, the classification performance of linear discriminant analysis classifier has been improved after data fusion. It is suggested that there has great potential as an alternative method for high-sensitive, rapid, label-free, economical diagnosis of bTBI.

Predictive Framework Development for User-Friendly On-Site Glucose Detection.

This study explores a smartphone-based spot detection framework for glucose in a rapid, simple, and affordable paper-based analytical device (PAD), which employs machine-learning algorithms to estimate various glucose concentrations. Herein, two different detection mixtures were chosen with chitosan (C) and without chitosan (WC) for the color change analysis. Being a biopolymer, chitosan improves the analytical performance of PADs when used with a chromogenic agent. Moreover, the influence of the illumination conditions and camera optics on the professed color of glucose strips was observed by choosing various illumination conditions and different smartphones. Hence, this study focuses on developing a framework for smartphone-based simple and user-friendly spot-based glucose detection (with a concentration range of 10-40 mM) at any illumination conditions and in any direction of illumination. Additionally, the combination of color spaces and machine-learning algorithms was applied for the signal enhancement. It was observed that the machine learning classifiers, cubic support vector machine (SVM) and narrow neutral network show higher accuracy for the WC samples, which are 92.7 and 92.3%, respectively. The samples with chitosan show higher accuracy for the linear discriminant and quadratic SVM classifiers, which are 94.1 and 93.9%, respectively. Simultaneously, cubic SVM shows ∼93% accuracy for both cases. In order to assess the performance of the devices, a blind test was also conducted. This study demonstrates the potential of the developed system for initial disease screening at the user end. By incorporating machine learning techniques, the platform can provide reliable and accurate results, thus paving the way for estimating the accuracy of the results for improved initial healthcare screening and diagnosis of any disease.

The potential of digital behavioural tests as a diagnostic aid for psychosis.

Timely interventions have a proven benefit for people experiencing psychotic illness. One bottleneck to accessing timely interventions is the referral process to the specialist team for early psychosis (STEP). Many general practitioners lack awareness or confidence in recognising psychotic symptoms or state. Additionally, referrals for people without apparent psychotic symptoms, although beneficial at a population level, lead to excessive workload for STEPs. There is a clear unmet need for accurate stratification of STEPs users and healthy cohorts. Here we propose a new approach to addressing this need via the application of digital behavioural tests. To demonstrate that digital behavioural tests can be used to discriminate between the STEPs users (SU; n = 32) and controls (n = 32, age and sex matched), we compared performance of five different classifiers applied to objective, quantitative and interpretable features derived from the 'mirror game' (MG) and trail making task (TMT). The MG is a movement coordination task shown to be a potential socio-motor biomarker of schizophrenia, while TMT is a neuropsychiatric test of cognitive function. All classifiers had AUC in the range of 0.84-0.92. The best of the five classifiers (linear discriminant classifier) achieved an outstanding performance, AUC = 0.92 (95%CI 0.75-1), Sensitivity = 0.75 (95%CI 0.5-1), Specificity = 1 (95%CI 0.75-1), evaluated on 25% hold-out and 1000 folds. Performance of all analysed classifiers is underpinned by the large effect sizes of the differences between the cohorts in terms of the features used for classification what ensures generalisability of the results. We also found that MG and TMT are unsuitable in isolation to successfully differentiate between SU with and without at-risk-mental-state or first episode psychosis with sufficient level of performance. Our findings show that standardised batteries of digital behavioural tests could benefit both clinical and research practice. Including digital behavioural tests into healthcare practice could allow precise phenotyping and stratification of the highly heterogenous population of people referred to STEPs resulting in quicker and more personalised diagnosis. Moreover, the high specificity of digital behavioural tests could facilitate the identification of more homogeneous clinical high-risk populations, benefiting research on prognostic instruments for psychosis. In summary, our study demonstrates that cheap off-the-shelf equipment (laptop computer and a leap motion sensor) can be used to record clinically relevant behavioural data that could be utilised in digital mental health applications.

Geographical origin identification of ginseng using near-infrared spectroscopy coupled with subspace-based ensemble classifiers

Ginseng is a well-known traditional herbal medicine and the ginseng available on the market may not actually be produced in a certain place as claimed. Traditional methods of identifying the geographical origin of Ginseng are subjective, time-consuming or destructive. A more efficient approach is desirable. The feasibility of combining near-infrared (NIR) spectroscopy with ensemble learning for discriminating ginseng producing area was explored. A total of 270 samples were collected and evenly partitioned into the training and test sets. Random subspace ensemble (RSE) that uses linear discriminant classifier (LDA) as weak learner (abbreviated RSE-LDA) was used to construct predictive models. Two parameters including the size of subspace and the number of learners in ensemble were optimized. Classic partial least algorithm (PLS) was applied to build the reference model. The sensitivity, specificity, and total accuracy of final RSE-LDA and PLS models were 97.8 %, 100 %, 99.3 %, and 93.3 %, 96.7 %, 95.6 %, respectively. In order to study the impact of training set composition on the results, the samples were randomly divided 200 times and the algorithm was run repeatedly to statistically analyze the sensitivity and specificity on the test set. Similar results were obtained. The effect of training set size was also investigated. It indicates that the combination of NIR spectroscopy with the RSE algorithm is a potential tool of discriminating the origin of Ginseng.

A Novel Event-Driven Spiking Convolutional Neural Network for Electromyography Pattern Recognition.

Electromyography (EMG) pattern recognition is an important technology for prosthesis control and human-computer interaction etc. However, the practical application of EMG pattern recognition is hampered by poor accuracy and robustness due to electrode shift caused by repeated wearing of the signal acquisition device. Moreover, the user's acceptability is low due to the heavy training burden, which is caused by the need for a large amount of training data by traditional methods. In order to explore the advantage of spiking neural network (SNN) in solving the poor robustness and heavy training burden problems in EMG pattern recognition, a spiking convolutional neural network (SCNN) composed of cyclic convolutional neural network (CNN) and fully connected modules is proposed and implemented in this study. High density surface electromyography (HD-sEMG) signals collected from 6 gestures of 10 subjects at 6 electrode positions are taken as the research object. Compared to CNN with the same structure, CNN-Long Short Term Memory (CNN-LSTM), linear kernel linear discriminant analysis classifier (LDA) and spiking multilayer perceptron (Spiking MLP), the accuracy of SCNN is 50.69%, 33.92%, 32.94% and 9.41% higher in the small sample training experiment, 6.50%, 4.23%, 28.73%, and 2.57% higher in the electrode shifts experiment respectively. In addition, the power consumption of SCNN is about 1/93 of CNN. The advantages of the proposed framework in alleviating user training burden, mitigating the adverse effect of electrode shifts and reducing power consumption make it very meaningful for promoting the development of user-friendly real-time myoelectric control system.