Intention Detection Research Articles

Adaptive neuro-fuzzy inference system model driven by the non-negative matrix factorization-extracted muscle synergy patterns to estimate lower limb joint movements

Background and objectiveFor patients with movement disorders, the main clinical focus is on exercise rehabilitation to help recover lost motor function, which is achieved by relevant assisted equipment. The basis for seamless control of the assisted equipment is to achieve accurate inference of the user's movement intentions in the human-machine interface. This study proposed a novel movement intention detection technology for estimating lower limb joint continuous kinematic variables following muscle synergy patterns, to develop applications for more efficient assisted rehabilitation training. MethodsThis study recruited 16 healthy males and 16 male patients with symptomatic patellar tendinopathy (VISA-P: 59.1 ± 8.7). The surface electromyography of 12 muscles and lower limb joint kinematic and kinetic data from healthy subjects and patients during step-off landings from 30 cm-high stair steps were collected. We subsequently solved the preprocessed data based on the established recursive model of second-order differential equation to obtain the muscle activation matrix, and then imported it into the non-negative matrix factorization model to obtain the muscle synergy matrix. Finally, the lower limb neuromuscular synergy pattern was then imported into the developed adaptive neuro-fuzzy inference system non-linear regression model to estimate the human movement intention during this movement pattern. ResultsSix muscle synergies were determined to construct the muscle synergy pattern driven ANFIS model. Three fuzzy rules were determined in most estimation cases. Combining the results of the four error indicators across the estimated variables indicates that the current model has excellent estimated performance in estimating lower limb joint movement. The estimation errors between the healthy (Angle: R2=0.98±0.03; Torque: R2=0.96±0.04) and patient (Angle: R2=0.98±0.02; Torque: R2=0.96±0.03) groups are consistent. ConclusionThe proposed model of this study can accurately and reliably estimate lower limb joint movements, and the effectiveness will also be radiated to the patient group. This revealed that our models also have certain advantages in the recognition of motor intentions in patients with relevant movement disorders. Future work from this study can be focused on sports rehabilitation in the clinical field by achieving more flexible and precise movement control of the lower limb assisted equipment to help the rehabilitation of patients.

A Myoelectric-based pattern recognition architecture for shoulder movement intent detection: toward a human-machine interface for transhumeral amputees

A Myoelectric-based pattern recognition architecture for shoulder movement intent detection: toward a human-machine interface for transhumeral amputees

Modeling and Analysis of Human Comfort in Human-Robot Collaboration.

The emergence and recent development of collaborative robots have introduced a safer and more efficient human-robot collaboration (HRC) manufacturing environment. Since the release of COBOTs, a great amount of research efforts have been focused on improving robot working efficiency, user safety, human intention detection, etc., while one significant factor-human comfort-has been frequently ignored. The comfort factor is critical to COBOT users due to its great impact on user acceptance. In previous studies, there is a lack of a mathematical-model-based approach to quantitatively describe and predict human comfort in HRC scenarios. Also, few studies have discussed the cases when multiple comfort factors take effect simultaneously. In this study, a multi-linear-regression-based general human comfort prediction model is proposed under human-robot collaboration scenarios, which is able to accurately predict the comfort levels of humans in multi-factor situations. The proposed method in this paper tackled these two gaps at the same time and also demonstrated the effectiveness of the approach with its high prediction accuracy. The overall average accuracy among all participants is 81.33%, while the overall maximum value is 88.94%, and the overall minimum value is 72.53%. The model uses subjective comfort rating feedback from human subjects as training and testing data. Experiments have been implemented, and the final results proved the effectiveness of the proposed approach in identifying human comfort levels in HRC.

Joint intent detection and slot filling for Turkish natural language understanding

Joint intent detection and slot filling for Turkish natural language understanding

Multimodal Deep Learning with Discriminant Descriptors for Offensive Memes Detection

A meme is a visual representation that illustrates a thought or concept. Memes are spreading steadily among people in this era of rapidly expanding social media platforms, and they are becoming increasingly popular forms of expression. In the domain of meme and emotion analysis, the detection of offensives is a crucial task. However, it can be difficult to identify and comprehend the underlying emotion of a meme because its content is multimodal. Additionally, there is a lack of memes datasets that address how offensive a meme is, and the existing ones in this context have a bias towards the dominant labels or categories, leading to an imbalanced training set. In this article, we present a descriptive balanced dataset to help detect the offensive nature of the meme’s content using a proposed multimodal deep learning model. Two deep semantic models, baseline BERT and hateXplain-BERT, are systematically combined with several deep ResNet architectures to estimate the severity of the offensive memes. This process is based on the Meme-Merge collection that we construct from two publicly available datasets. The experimental results demonstrate the model’s effectiveness in classifying offensive memes, achieving F1 scores of 0.7315 and 0.7140 for the baseline datasets and Meme-Merge, respectively. The proposed multimodal deep learning approach also outperformed the baseline model in three meme tasks: metaphor understanding, sentiment understanding, and intention detection.

CTRAN: CNN-Transformer-based network for natural language understanding

Intent-detection (ID) and slot-filling (SF) are fundamental tasks for natural language understanding. This study introduces a new encoder–decoder CNN-Transformer-based architecture (CTRAN) designed for ID and SF. The encoder integrates of BERT, followed by several convolutional layers with different kernel sizes. We propose using a kernel size of 1 to preserve the one-to-one correspondence between input tokens and output tags. Subsequently, we rearrange the output of the convolutional layer using the window feature sequence and apply stacked Transformer encoders. The ID decoder leverages self-attention and a linear layer, while the SF decoder employs an aligned Transformer decoder with a zero diagonal mask, facilitating alignment between output tags and input tokens. We evaluate our model on ATIS and SNIPS datasets, achieving 98.46% and 98.30% F1 score for the SF task, respectively. These results surpass the previous state-of-the-art by 0.64% and 0.99%. Moreover, we compare two strategies: using a language model as the encoder or as word embedding. We find out that the latter strategy yields better results.

Task-oriented human-robot interaction control of a robotic glove utilizing forearm electromyography

Task-oriented myoelectric assistive control of a robotic glove based on forearm electromyography for practical grasping tasks is investigated in this work. Three grasping functions are considered: human intent detection of grasping tasks, grasping mode recognition, and grasping force control. Different combinations of forearm electromyographic signals are adopted for the three functions. Firstly, the overall electromyographic signal is used to trigger the whole grasping task. Secondly, a novel Long Short-Term Memory network is utilized to classify various grasping modes, including pinch grasping and palmar grasping, by analyzing eight-channel electromyographic signals. Thirdly, two-channel proportional control of grasping force using electromyographic flexor/extensor signals is adopted for the robotic glove such that the human hand can relax during the grasping task, while the robotic glove maintains the grasping force. To this end, finite state machine based hierarchical control architecture is proposed for the whole grasping task. Experiments are conducted to validate the proposed task-oriented assistive control method, and the results clearly demonstrates the potential of the proposed method in rehabilitation therapy.

Joint Modelling of Slot Filling and Intent Detection in Constrained Resource Scenarios

In the era of pervasive smart devices, natural language understanding (NLU) holds a pivotal role for facilitating intelligent interactions and decision-making. Core to NLU are slot filling and intent recognition, essential tasks for comprehending user input. While joint modelling of these tasks has gained prominence, the challenges of realizing efficient joint models on resource-constrained devices have emerged as significant. These devices possess limited computational capacity and real-time requirements, necessitating lightweight and efficient models. In this study, we explore the design of a resource-efficient joint model for slot filling and intent recognition. Through leveraging BERT, , graph neural networks, and mask mechanisms, our model achieves the dual goals of semantic slot prediction and intent classification. We focus on model design, training, and real-time inference, aiming to contribute to the paradigm of resource-constrained natural language understanding. Our investigation demonstrates the efficacy of our approach, even when working with a reduced dataset, underscoring the model's applicability to real-world scenarios with limited resources.

Machine Learning for Human Motion Intention Detection.

The gait pattern of exoskeleton control conflicting with the human operator's (the pilot) intention may cause awkward maneuvering or even injury. Therefore, it has been the focus of many studies to help decide the proper gait operation. However, the timing for the recognization plays a crucial role in the operation. The delayed detection of the pilot's intent can be equally undesirable to the exoskeleton operation. Instead of recognizing the motion, this study examines the possibility of identifying the transition between gaits to achieve in-time detection. This study used the data from IMU sensors for future mobile applications. Furthermore, we tested using two machine learning networks: a linearfFeedforward neural network and a long short-term memory network. The gait data are from five subjects for training and testing. The study results show that: 1. The network can successfully separate the transition period from the motion periods. 2. The detection of gait change from walking to sitting can be as fast as 0.17 s, which is adequate for future control applications. However, detecting the transition from standing to walking can take as long as 1.2 s. 3. This study also find that the network trained for one person can also detect movement changes for different persons without deteriorating the performance.

Feasibility of force myography for the direct control of an assistive robotic hand orthosis in non-impaired individuals

BackgroundAssistive robotic hand orthoses can support people with sensorimotor hand impairment in many activities of daily living and therefore help to regain independence. However, in order for the users to fully benefit from the functionalities of such devices, a safe and reliable way to detect their movement intention for device control is crucial. Gesture recognition based on force myography measuring volumetric changes in the muscles during contraction has been previously shown to be a viable and easy to implement strategy to control hand prostheses. Whether this approach could be efficiently applied to intuitively control an assistive robotic hand orthosis remains to be investigated.MethodsIn this work, we assessed the feasibility of using force myography measured from the forearm to control a robotic hand orthosis worn on the hand ipsilateral to the measurement site. In ten neurologically-intact participants wearing a robotic hand orthosis, we collected data for four gestures trained in nine arm configurations, i.e., seven static positions and two dynamic movements, corresponding to typical activities of daily living conditions. In an offline analysis, we determined classification accuracies for two binary classifiers (one for opening and one for closing) and further assessed the impact of individual training arm configurations on the overall performance.ResultsWe achieved an overall classification accuracy of 92.9% (averaged over two binary classifiers, individual accuracies 95.5% and 90.3%, respectively) but found a large variation in performance between participants, ranging from 75.4 up to 100%. Averaged inference times per sample were measured below 0.15 ms. Further, we found that the number of training arm configurations could be reduced from nine to six without notably decreasing classification performance.ConclusionThe results of this work support the general feasibility of using force myography as an intuitive intention detection strategy for a robotic hand orthosis. Further, the findings also generated valuable insights into challenges and potential ways to overcome them in view of applying such technologies for assisting people with sensorimotor hand impairment during activities of daily living.

When searching helps you see: Bridging the gap between incidental and intentional change detection.

Participants in incidental change detection studies often miss large changes to visually salient or conceptually relevant objects such as actor substitutions across video cuts, but there are competing explanations of why participants fail to detect these changes. According to an integrative processing account, object-based attention typically induces integrated representation and comparison processes sufficient to detect changes to that object. On this view, participants miss changes in incidental paradigms because those paradigms fail to elicit the level of attention necessary to trigger integrated representation and comparison processes. In contrast, a selective processing account posits that representation and comparison processes needed to detect changes do not occur by default, even for attended objects, but are only elicited in response to specific functional needs. In four experiments, we tested detection of actor substitutions when participants engaged in tasks that required actor identity processing but did not necessarily require the combination of processes necessary to detect changes. Change blindness for actor substitutions persisted when participants counted the number of actors in the video and sometimes persisted when participants were instructed to remember the substituted actor for later recall. Change blindness consistently diminished, however, when participants were shown the prechange actor before or during the video and instructed to search for that actor in the video. Our results refine the contrast between selective and integrative processing by specifying how task demands to create durable visual representations can remain independent of comparison processes, while search demands can induce integrative comparison processes in a naturalistic setting. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Error-related Potentials in a Virtual Pick-and-Place Experiment: Toward Real-world Shared-control.

In Human-Robot Collaboration setting a robot may be controlled by a user directly or through a Brain-Computer Interface that detects user intention, and it may act as an autonomous agent. As such interaction increases in complexity, conflicts become inevitable. Goal conflicts can arise from different sources, for instance, interface mistakes - related to misinterpretation of human's intention - or errors of the autonomous system to address task and human's expectations. Such conflicts evoke different spontaneous responses in the human's brain, which could be used to regulate intrinsic task parameters and to improve system response to errors - leading to improved transparency, performance, and safety. To study the possibility of detecting interface and agent errors, we designed a virtual pick and place task with sequential human and robot responsibility and recorded the electroencephalography (EEG) activity of six participants. In the virtual environment, the robot received a command from the participants through a computer keyboard or it moved as autonomous agent. In both cases, artificial errors were defined to occur in 20% - 25% of the trials. We found differences in the responses to interface and agent errors. From the EEG data, correct trials, interface errors, and agent errors were truly predicted for 51.62% ± 9.99% (chance level 38.21%) of the pick movements and 46.84%±6.62% (chance level 36.99%) for the place movements in a pseudo-asynchronous fashion. Our study suggests that in a human-robot collaboration setting one may improve the future performance of a system with intention detection and autonomous modes. Specific examples could be Neural Interfaces that replace and restore motor functions.

Development of a Bioimpedance and sEMG Fusion Sensor for Gait Phase Detection: Validation with a Transtibial Amputee.

Accurate gait phase detection is crucial for safe and efficient robotic prosthesis control in lower limb amputees. Several sensing modalities, including mechanical and biological signals, have been proposed to improve the accuracy of gait phase detection. In this paper, we propose a bioimpedance and sEMG fusion sensor for high-accuracy gait phase detection. We fabricated a wearable band-type sensor for multichannel bioimpedance and sEMG measurement, and we conducted gait experiments with a transtibial amputee to obtain biosignal data. Finally, we trained a deep-learning-based gait phase detection algorithm and evaluated its detection performance. Our results showed that using both bioimpedance and sEMG yielded the highest accuracy of 95.1%. Using only sEMG yielded a higher accuracy (90.9%) than that using only bioimpedance (85.1%). Therefore, we conclude that using both signals simultaneously is beneficial for improving the accuracy of gait phase detection. In addition, the proposed sensor can be applied to several applications by improving the accuracy of motion intention detection.

Decoupling brain activations of muscle-caused activations and mental intention-cause activations using the general linear model: A functional near-infrared spectroscopy study.

Integrating a brain-computer interface into a lower-limb medical rehabilitation assistive device can enhance rehabilitation efficiency. The latest research in the field focuses on the decoding performance of different motions. However, the difference between muscle-caused primitive activation and mental intention-caused activation has not been fully investigated. Thus, our study tried to decouple these two kinds of cerebral activation using a general linear model (GLM). Nine healthy and right-handed subjects were recruited for a two-section experiment. They were asked to extend or flex their knees while seated in the first section of the experiment or standing in the second section. Functional near-infrared spectroscopy (fNIR) was adopted to monitor their hemodynamic changes. Two groups of paradigms (one for circle-wise analysis, the other for full-section analysis) were constructed from the experimental paradigm. Each group consisted of three (the first intention, the second intention, and the muscle activation). The constructed paradigms were fed to the Balloon model for six desired hemodynamic responses (dHRFs). The regressor of GLM consisted of three dHRFs and the corresponding motion artifacts and drifts. The simulated physiological noises were included in the structured background matrix. The results showed that all subjects had similar cerebral activation patterns for the intention to extend or flex knees. The activation during musclecaused activation was less intense than that caused by both intentions. This finding can help further research on more efficient motion intention detection and the possibility of multiple motions decoding.

Humans perceive warmth and competence in artificial intelligence

Artificial intelligence (A.I.) increasingly suffuses everyday life. However, people are frequently reluctant to interact with A.I. systems. This challenges both the deployment of beneficial A.I. technology and the development of deep learning systems that depend on humans for oversight, direction, and regulation. Nine studies (N= 3,300) demonstrate that social-cognitive processes guide human interactions across a diverse range of real-world A.I. systems. Across studies, perceived warmth and competence emerge prominently in participants' impressions of A.I. systems. Judgments of warmth and competence systematically depend on human-A.I. interdependence and autonomy. In particular, participants perceive systems that optimize interests aligned with human interests as warmer and systems that operate independently from human direction as more competent. Finally, a prisoner's dilemma game shows that warmth and competence judgments predict participants' willingness to cooperate with a deep-learning system. These results underscore the generality of intent detection to perceptions of a broad array of algorithmic actors.

EEG-based emergency braking intention detection during simulated driving

BackgroundCurrent research related to electroencephalogram (EEG)-based driver’s emergency braking intention detection focuses on recognizing emergency braking from normal driving, with little attention to differentiating emergency braking from normal braking. Moreover, the classification algorithms used are mainly traditional machine learning methods, and the inputs to the algorithms are manually extracted features.MethodsTo this end, a novel EEG-based driver’s emergency braking intention detection strategy is proposed in this paper. The experiment was conducted on a simulated driving platform with three different scenarios: normal driving, normal braking and emergency braking. We compared and analyzed the EEG feature maps of the two braking modes, and explored the use of traditional methods, Riemannian geometry-based methods, and deep learning-based methods to predict the emergency braking intention, all using the raw EEG signals rather than manually extracted features as input.ResultsWe recruited 10 subjects for the experiment and used the area under the receiver operating characteristic curve (AUC) and F1 score as evaluation metrics. The results showed that both the Riemannian geometry-based method and the deep learning-based method outperform the traditional method. At 200 ms before the start of real braking, the AUC and F1 score of the deep learning-based EEGNet algorithm were 0.94 and 0.65 for emergency braking vs. normal driving, and 0.91 and 0.85 for emergency braking vs. normal braking, respectively. The EEG feature maps also showed a significant difference between emergency braking and normal braking. Overall, based on EEG signals, it was feasible to detect emergency braking from normal driving and normal braking.ConclusionsThe study provides a user-centered framework for human–vehicle co-driving. If the driver's intention to brake in an emergency can be accurately identified, the vehicle's automatic braking system can be activated hundreds of milliseconds earlier than the driver's real braking action, potentially avoiding some serious collisions.

Novel Intent Detection and Active Learning Based Classification (Student Abstract)

Novel intent class detection is an important problem in real world scenario for conversational agents for continuous interaction. Several research works have been done to detect novel intents in a mono-lingual (primarily English) texts and images. But, current systems lack an end-to-end universal framework to detect novel intents across various different languages with less human annotation effort for mis-classified and system rejected samples. This paper proposes NIDAL (Novel Intent Detection and Active Learning based classification), a semi-supervised framework to detect novel intents while reducing human annotation cost. Empirical results on various benchmark datasets demonstrate that this system outperforms the baseline methods by more than 10% margin for accuracy and macro-F1. The system achieves this while maintaining overall annotation cost to be just ~6-10% of the unlabeled data available to the system.

Learning to Select from Multiple Options

Many NLP tasks can be regarded as a selection problem from a set of options, such as classification tasks, multi-choice question answering, etc. Textual entailment (TE) has been shown as the state-of-the-art (SOTA) approach to dealing with those selection problems. TE treats input texts as premises (P), options as hypotheses (H), then handles the selection problem by modeling (P, H) pairwise. Two limitations: first, the pairwise modeling is unaware of other options, which is less intuitive since humans often determine the best options by comparing competing candidates; second, the inference process of pairwise TE is time-consuming, especially when the option space is large. To deal with the two issues, this work first proposes a contextualized TE model (Context-TE) by appending other k options as the context of the current (P, H) modeling. Context-TE is able to learn more reliable decision for the H since it considers various context. Second, we speed up Context-TE by coming up with Parallel-TE, which learns the decisions of multiple options simultaneously. Parallel-TE significantly improves the inference speed while keeping comparable performance with Context-TE. Our methods are evaluated on three tasks (ultra-fine entity typing, intent detection and multi-choice QA) that are typical selection problems with different sizes of options. Experiments show our models set new SOTA performance; particularly, Parallel-TE is faster than the pairwise TE by k times in inference.

A Scope Sensitive and Result Attentive Model for Multi-Intent Spoken Language Understanding

Multi-Intent Spoken Language Understanding (SLU), a novel and more complex scenario of SLU, is attracting increasing attention. Unlike traditional SLU, each intent in this scenario has its specific scope. Semantic information outside the scope even hinders the prediction, which tremendously increases the difficulty of intent detection. More seriously, guiding slot filling with these inaccurate intent labels suffers error propagation problems, resulting in unsatisfied overall performance. To solve these challenges, in this paper, we propose a novel Scope-Sensitive Result Attention Network (SSRAN) based on Transformer, which contains a Scope Recognizer (SR) and a Result Attention Network (RAN). SR assignments scope information to each token, reducing the distraction of out-of-scope tokens. RAN effectively utilizes the bidirectional interaction between SF and ID results, mitigating the error propagation problem. Experiments on two public datasets indicate that our model significantly improves SLU performance (5.4% and 2.1% on Overall accuracy) over the state-of-the-art baseline.

Development of Early Detection Instruments of Building Intention in Elementary School Students

Bullying in childhood is very detrimental and has a negative impact both in the short and long term. Bullying can cause stress, decreased self-esteem, difficulty getting along, causing psychosocial problems such as depression, loneliness, and anxiety, triggering somatic symptoms, low academic achievement, dropping out of school, increasing psychosis, and encouraging suicidal ideation. Bullying must be prevented and reduced through various efforts such as bullying prevention programs, namely detection and intervention against bullying. In order to detect the potential for bullying to occur, this research focuses on producing instruments for the early detection of child bullying intentions in elementary school students. The instrument developed was an early detection questionnaire on child bullying intentions (DDIPA). The study involved 1260 elementary school students in the early grades. Data analysis using Rasch modeling. The result is a DDIPA instrument consisting of 39 valid and reliable statement items. The DDIPA instrument can be used by researchers, teachers, and counselors in order to detect potential bullying in schools.