Intention Detection

This research developed a wearable lower extremity assistive device intended to aid stroke patient during rehabilitation. The device specifically aims to assist the patient in sit-to-stand, stand-to-sit, and level-walking tasks in order to promote active gait rehabilitation exercises. The device adopts an anthropomorphic structure with hip and knee joint actuated in sagittal plane. A finite state machine strategy was proposed to control the device. At different states, appropriate assist torque is added to each joint. EMG signals are used to assess the assist performance. Tests on an able-bodied subject show that the device could successfully detect and transit between states. In sit-to-stand tasks, the integrated EMG (iEMG) of the Vastus Medialis for standing up with 11 Nm of assistance torque were found to be significantly less (P = 0.00187) than the iEMG of without assistance for standing up which indicate reduced muscle effort with the device assistance. Results show the device could potentially assist stroke patient in similar tasks.

This research developed a wearable lower extremity assistive device intended to aid stroke patient during rehabilitation. The device specifically aims to assist the patient in sit-to-stand, stand-to-sit, and level-walking tasks in order to promote active gait rehabilitation exercises. The device adopts an anthropomorphic structure with hip and knee joint actuated in sagittal plane. A finite state machine strategy was proposed to control the device. At different states, appropriate assist torque is added to each joint. EMG signals are used to assess the assist performance. Tests on an able-bodied subject show that the device could successfully detect and transit between states. In sit-to-stand tasks, the integrated EMG (iEMG) of the Vastus Medialis for standing up with 11 Nm of assistance torque were found to be significantly less (P = 0.00187) than the iEMG of without assistance for standing up which indicate reduced muscle effort with the device assistance. Results show the device could potentially assist stroke patient in similar tasks.

Great strides have been taken in advancing the domain of Natural Language Processing(NLP) over the past few years. The introduction of attention mechanism in neural networks has led to the creation very complex and efficient architectures like the BERT and GPT. These architectures have been well established as the first choice for most problem statements but, there application in certain simpler use cases generally undermines their real potential. The use case I explore in this research paper is the Natural Language Processing with Disaster Tweets. This is a binary classification task for picking out tweets that refer to a disaster. In this research paper I report the performance of Term Frequency based algorithm without leveraging any vector embedding techniques. Since I am making use of an open source dataset from Kaggle, I have received a leaderboard score of 0.78. The scoring metric used for the leaderboard is the F1 metric and the best score stands at 0.85. The approach leverages the ability of a inverse document frequency(IDF) metric to readily eliminate "stop word" from a corpus and uses a modified term frequency metric<put definition reference> to classify the sentences. The said term frequency metric has been designed in a way to eliminate any impact of stop words that might have escaped from the IDF filter. This paper presents the detailed description of all the algorithms and metrics utilized before coming to a final conclusion.

In this paper, we present a multi-modal perception based framework to realize a nonintrusive domestic assistive robotic system. It is non-intrusive in that it only starts interaction with a user when it detects the user's intention to do so. All the robot's actions are based on multi-modal perceptions which include user detection based on RGB-D data, user's intention-for-interaction detection with RGB-D and audio data, and communication via user distance mediated speech recognition. The utilization of multi-modal cues in different parts of the robotic activity paves the way to successful robotic runs (94% success rate). Each presented perceptual component is systematically evaluated using appropriate dataset and evaluation metrics. Finally the complete system is fully integrated on the PR2 robotic platform and validated through system sanity check runs and user studies with the help of 17 volunteer elderly participants.

Understanding people's intention, be it action or thought, plays a fundamental role in establishing coherent communication amongst people, especially in non-proactive robotics, where the robot has to understand explicitly when to start an interaction in a natural way. In this work, a novel approach is presented to detect people's intention-for-interaction. The proposed detector fuses multimodal cues, including estimated head pose, shoulder orientation and vocal activity detection, using a probabilistic discrete state Hidden Markov Model. The multimodal detector achieves up to 80% correct detection rates improving purely audio and RGB-D based variants.

Dentro del reconocimiento automático del habla, los modelos de lenguaje estadísticos basados en la probabilidad de secuencia de palabras (n-gramas) suponen uno de los dos pilares sobre los que se basa su correcto funcionamiento. En este trabajo se expone el impacto que tienen sobre las prestaciones de reconocimiento a medida que estos modelos se mejoran con más texto de mejor calidad, cuando estos se ajustan a la aplicación final del sistema, y por lo tanto, cuando se reducen el número de palabras fuera de vocabulario (Out Of Vocabulary - OOV). El reconocedor con los distintos modelos de lenguaje ha sido aplicado sobre cortes de audio correspondientes a tres marcos experimentales: oralidad formal, habla en noticiarios, y TED talks en gallego. Los resultados obtenidos muestran claramente una mejora sobre los marcos experimentales propuestos.Within the automatic speech recognition, statistical language models based on the probability of word sequences (n-grams) represent one of the tw...

Resumen: El proyecto INVOCA consiste en un sistema de comprensión de comunicaciones habladas tierra-aire para el control de tráfico aéreo. Este sistema emplea una arquitectura de dos reconocedores de habla en paralelo con un módulo de detección de idioma (para inglés y castellano). La salida del reconocedor pasa a un modulo de comprensión basado en reglas dependientes de contexto que extrae los conceptos clave de cada tarea.

Se presenta un sistema de comprensión de comunicaciones habladas en dos idiomas, castellano e inglés, para el control de tráfico aéreo. Se emplea una arquitectura con dos reconocedores en paralelo más un módulo de detección de idioma. La salida del reconocedor en el idioma elegido pasa al sistema de comprensión basado en reglas dependientes de contexto que extrae los conceptos clave. Palabras clave: Reconocimiento, multi-idioma, modelado de lenguaje, comprensión, control aéreo.

Resumen: El proyecto INVOCA consiste en un sistema de comprensión de comunicaciones habladas tierra-aire para el control de tráfico aéreo. Este sistema emplea una arquitectura de dos reconocedores de habla en paralelo con un módulo de detección de idioma (para inglés y castellano). La salida del reconocedor pasa a un modulo de comprensión basado en reglas dependientes de contexto que extrae los conceptos clave de cada tarea.

Resumen: El proyecto INVOCA consiste en un sistema de comprensión de comunicaciones habladas tierra-aire para el control de tráfico aéreo. Este sistema emplea una arquitectura de dos reconocedores de habla en paralelo con un módulo de detección de idioma (para inglés y castellano). La salida del reconocedor pasa a un modulo de comprensión basado en reglas dependientes de contexto que extrae los conceptos clave de cada tarea.

dictionary of specific phrases from the domain of the application. The algorithm proposed here, despite being designed for very specific domains, is of general application. That is, the phrases that must be recognized can be defined specifically for each application, without the algorithm having to be modified. It is enough to indicate to the algorithm the set of sentences on which it must work. The complexity of the algorithm is O(tn), where t is the number of words contained in the transcript to be corrected and n is the number of phrases specific of the domain.

Great strides have been taken in advancing the domain of Natural Language Processing(NLP) over the past few years. The introduction of attention mechanism in neural networks has led to the creation very complex and efficient architectures like the BERT and GPT. These architectures have been well established as the first choice for most problem statements but, there application in certain simpler use cases generally undermines their real potential. The use case I explore in this research paper is the Natural Language Processing with Disaster Tweets. This is a binary classification task for picking out tweets that refer to a disaster. In this research paper I report the performance of Term Frequency based algorithm without leveraging any vector embedding techniques. Since I am making use of an open source dataset from Kaggle, I have received a leaderboard score of 0.78. The scoring metric used for the leaderboard is the F1 metric and the best score stands at 0.85. The approach leverages the ability of a inverse document frequency(IDF) metric to readily eliminate "stop word" from a corpus and uses a modified term frequency metric<put definition reference> to classify the sentences. The said term frequency metric has been designed in a way to eliminate any impact of stop words that might have escaped from the IDF filter. This paper presents the detailed description of all the algorithms and metrics utilized before coming to a final conclusion.

In human-computer collaboration involving haptics, a key issue that remains to be solved is to establish an intuitive communication between the partners. Even though computers are widely used to aid human operators in teleoperation, guidance, and training, since they lack the adaptability, versatility, and awareness of a human, their ability to improve efficiency and effectiveness in dynamic tasks is limited. We suggest that the communication between a human and a computer can be improved if it involves a decision making process in which the computer is programmed to infer the intentions of the human operator and dynamically adjust the control levels of the interacting parties to facilitate a more intuitive interaction setup. In this paper, we investigate the utility of such a dynamic role exchange mechanism where partners negotiate through the haptic channel to trade their control levels on a collaborative task. We examine the energy consumption, the work done on the manipulated object, and the joint efficiency in addition to the task performance. We show that when compared to an equal control condition, a role exchange mechanism improves task performance and the joint efficiency of the partners. We also show that augmenting the system with additional informative visual and vibrotactile cues, which are used to display the state of interaction, allows the users to become aware of the underlying role exchange mechanism and utilize it in favor of the task. These cues also improve the user's sense of interaction and reinforce his/her belief that the computer aids with the execution of the task.

Since the strict separation of working spaces of humans and robots has experienced a softening due to recent robotics research achievements, close interaction of humans and robots comes rapidly into reach. In this context, physical human–robot interaction raises a number of questions regarding a desired intuitive robot behavior. The continuous bilateral information and energy exchange requires an appropriate continuous robot feedback. Investigating a cooperative manipulation task, the desired behavior is a combination of an urge to fulfill the task, a smooth instant reactive behavior to human force inputs and an assignment of the task effort to the cooperating agents. In this paper, a formal analysis of human–robot cooperative load transport is presented. Three different possibilities for the assignment of task effort are proposed. Two proposed dynamic role exchange mechanisms adjust the robot’s urge to complete the task based on the human feedback. For comparison, a static role all...

Great strides have been taken in advancing the domain of Natural Language Processing(NLP) over the past few years. The introduction of attention mechanism in neural networks has led to the creation very complex and efficient architectures like the BERT and GPT. These architectures have been well established as the first choice for most problem statements but, there application in certain simpler use cases generally undermines their real potential. The use case I explore in this research paper is the Natural Language Processing with Disaster Tweets. This is a binary classification task for picking out tweets that refer to a disaster. In this research paper I report the performance of Term Frequency based algorithm without leveraging any vector embedding techniques. Since I am making use of an open source dataset from Kaggle, I have received a leaderboard score of 0.78. The scoring metric used for the leaderboard is the F1 metric and the best score stands at 0.85. The approach leverages the ability of a inverse document frequency(IDF) metric to readily eliminate "stop word" from a corpus and uses a modified term frequency metric<put definition reference> to classify the sentences. The said term frequency metric has been designed in a way to eliminate any impact of stop words that might have escaped from the IDF filter. This paper presents the detailed description of all the algorithms and metrics utilized before coming to a final conclusion.

Resumen: Uno de los aspectos más complejos de cualquier sistema de diálogo consiste en identificar la intención del usuario y en distinguir los errores de reconocimiento. En este artículo presentamos una nueva aproximación a la tarea de decidir qué ha dicho el usuario y qué quiere hacer en cada momento del diálogo. Para ello empleamos clasificadores estadísticos dirigidos por medidas de confianza generadas en paralelo por varios reconocedores adaptados a los diferentes temas del diálogo. Esta aproximación se ha mostrado especialmente adecuada para temas difíciles, como los nombres propios o las confirmaciones. La arquitectura utilizada mejora sustancialmente la tasa de reconocimiento y permite identificar la intención del usuario y detectar los cambios de tema. Palabras clave: Sistemas de diálogo, medidas de confianza, verificación de tema, detección de intención, reconocimiento de voz.

espanolEsta investigacion aborda el estudio de la vida y obra del escultor Juan Gonzalez Moreno, murciano de nacimiento, que desarrollo gran parte de su actividad escultorica en la etapa franquista, favoreciendo con su produccion artistica, la consolidacion del imaginario politico-religioso del regimen dictatorial. Su formacion se desarrollo, inicialmente, en los talleres de la ciudad de Murcia, trasladandose tiempo despues a Madrid, donde tomo contacto con las vanguardias de la epoca e hizo amistad con otros escultores del momento, como Benjamin Palencia y Alberto Sanchez. Al iniciarse la Guerra Civil volvio a Murcia, colaborando con el bando republicano como integrante de la Comision Municipal que se creo para salvaguardar el patrimonio artistico y religioso de la capital de la Region, circunstancia que, una vez finalizado el conflicto belico, no solo no le supuso depuracion profesional, sino que al contrario, tuvo el privilegio de poder salir del pais para continuar su formacion ...

Many applications benefit from the cooperation between humans and robots. This paper presents a summary of our efforts to investigate the necessary components that en-hance various features of interaction between humans and robots in highly dynamic joint manipulation tasks. We sug-gest that the communication between a human and a robot would benefit from a decision making process in which the robot can dynamically adjust its control level during the task based on the intentions of the human. In a virtual en-vironment, we allow the human to interact dynamically and smoothly with a computer agent (i.e. a robot). We define leader and follower roles for the partners, and using a role exchange mechanism, we enable the partners to negotiate through force information to exchange roles. We show that when compared to the equal control condition, the role ex-change mechanism improves task performance and the joint efficiency of the partners. We also show that users bene-fit from additional in...

This dissertation aims to present a perspective to build more natural shared control systems for physical human-robot cooperation. As the tasks become more complex and more dynamic, many shared control schemes fail to meet the expectation of an effortless interaction that resembles human-human sensory communication. Since such systems are mainly built to improve task performance, the richness of sensory communication is of secondary concern. We suggest that effective cooperation can be achieved when the human’s and the robot’s roles within the task are dynamically updated during the execution of the task. These roles define states for the system, in which the robot’s control leads or follows the human’s actions. In such a system, a state transition can occur at certain times if the robot can determine the user’s intention for gaining/relinquishing control. Specifically, with these state transitions we assign certain roles to the human and the robot. We believe that only by employing...

This paper presents an approach for early recognition of gait initiation and termination using wearable inertial measurement units and pressuresensitive insoles. Body joint angles, ground reaction force and center of plantar pressure of each foot are obtained from these sensors and input into a supervised learning algorithm. For gait initiation, the algorithm detects two events: gait onset (the first detectable change from the baseline state) and toe-off. For gait termination, the algorithm segments gait into different steps, measures the signals over a window at the beginning of each step, and determines whether the measurement belongs to the final step. The approach is validated with 10 subjects at two different gait speeds, with both within-subject and subject-independent crossvalidation. Results show that the inertial measurement units are generally more useful than insoles during both gait initiation and termination, though combining both types of sensors results in better onset detection and easier segmentation of gait into different steps. However, for best performance the algorithms should be trained for each subject separately, and the gait termination recognition algorithm is not very robust with regard to gait speed.

Backward walking (BW) is being increasingly used in neurologic and orthopedic rehabilitation as well as in sports to promote balance control as it provides a unique challenge to the sensorimotor control system. The identification of initial foot contact (IC) and terminal foot contact (TC) events is crucial for gait analysis. Data of optical motion capture (OMC) kinematics and inertial motion units (IMUs) are commonly used to detect gait events during forward walking (FW). However, the agreement between such methods during BW has not been investigated. In this study, the OMC kinematics and inertial data of 10 healthy young adults were recorded during BW and FW on a treadmill at different speeds. Gait events were measured using both kinematics and inertial data and then evaluated for agreement. Excellent reliability (Interclass Correlation &gt; 0.9) was achieved for the identification of both IC and TC. The absolute differences between methods during BW were 18.5 ± 18.3 and 20.4 ± 15....

We investigate how collaborative guidance can be realized in multi-modal virtual environments for dynamic tasks involving motor control. Haptic guidance in our context can be defined as any form of force/tactile feedback that the computer generates to help a user execute a task in a faster, more accurate, and subjectively more pleasing fashion. In particular, we are interested in determining guidance mechanisms that best facilitate task performance and arouse a natural sense of collaboration. We suggest that a haptic guidance system can be further improved if it is supplemented with a role exchange mechanism, which allows the computer to adjust the forces it applies to the user in response to his/her actions. Recent work on collaboration and role exchange presented new perspectives on defining roles and interaction. However existing approaches mainly focus on relatively basic environments where the state of the system can be defined with a few parameters. We designed and implemented a complex and highly dynamic multimodal game for testing our interaction model. Since the state space of our application is complex, role exchange needs to be implemented carefully. We defined a novel negotiation process, which facilitates dynamic communication between the user and the computer, and realizes the exchange of roles using a three-state finite state machine. Our preliminary results indicate that even though the negotiation and role exchange mechanism we adopted does not improve performance in every evaluation criteria, it introduces a more personal and human-like interaction model.

Since the strict separation of working spaces of humans and robots has experienced a softening due to recent robotics research achievements, close interaction of humans and robots comes rapidly into reach. In this context, physical human-robot interaction raises a number of questions regarding a desired intuitive robot behavior. The continuous bilateral information and energy exchange requires an appropriate continuous robot feedback. Investigating a cooperative manipulation task, the desired behavior is a combination of an urge to fulfill the task, a smooth instant reactive behavior to human force inputs and an assignment of the task effort to the cooperating agents. In this paper, a formal analysis of human-robot cooperative load transport is presented. Three different possibilities for the assignment of task effort are proposed. Two proposed dynamic role exchange mechanisms adjust the robot's urge to complete the task based on the human feedback. For comparison, a static role allocation strategy not relying on the human agreement feedback is investigated as well. All three role allocation mechanisms are evaluated in a user study that involves large-scale kinesthetic interaction and full-body human motion. Results show tradeoffs between subjective and objective performance measures stating a clear objective advantage of the proposed dynamic role allocation scheme.

In human-computer collaboration involving haptics, a key issue that remains to be solved is to establish an intuitive communication between the partners. Even though computers are widely used to aid human operators in teleoperation, guidance, and training, since they lack the adaptability, versatility, and awareness of a human, their ability to improve efficiency and effectiveness in dynamic tasks is limited. We suggest that the communication between a human and a computer can be improved if it involves a decision making process in which the computer is programmed to infer the intentions of the human operator and dynamically adjust the control levels of the interacting parties to facilitate a more intuitive interaction setup. In this paper, we investigate the utility of such a dynamic role exchange mechanism where partners negotiate through the haptic channel to trade their control levels on a collaborative task. We examine the energy consumption, the work done on the manipulated object, and the joint efficiency in addition to the task performance. We show that when compared to an equal control condition, a role exchange mechanism improves task performance and the joint efficiency of the partners. We also show that augmenting the system with additional informative visual and vibrotactile cues, which are used to display the state of interaction, allows the users to become aware of the underlying role exchange mechanism and utilize it in favor of the task. These cues also improve the user's sense of interaction and reinforce his/her belief that the computer aids with the execution of the task.

This paper presents a study of gait intention detection using force plates, inertial measurement units and an optical measurement system. The main goal is to detect gait initiation before heel-off and toe-off. Several established supervised machine learning methods are used to detect the onset of gait initiation, the first heel-off and the first toe-off. Events manually annotated by an expert serve as a reference. Results show that force plate signals are the most useful sensor, allowing gait onset to be detected with a mean absolute error of 0.12 seconds. Inertial measurement units are less accurate, with a mean absolute error for gait onset detection of 0.22 seconds. However, the decreased accuracy is primarily due to a small number of poorly detected outliers. The accuracy of the different supervised methods is also compared. For practical use, we recommend a combination of inertial measurement units and in-shoe pressure sensors, with different supervised methods used to detect different events.

Previous studies have presented algorithms for detection of turns during gait using wearable sensors, but those algorithms were not built for real-time use. This paper therefore investigates the optimal approach for real-time detection of planned turns during gait using wearable inertial measurement units. Several different sensor positions (head, back and legs) and three different detection criteria (orientation, angular velocity and both) are compared with regard to their ability to correctly detect turn onset. Furthermore, the different sensor positions are compared with regard to their ability to predict the turn direction and amplitude. The evaluation was performed on ten healthy subjects who performed left/right turns at three amplitudes (22, 45 and 90 degrees). Results showed that turn onset can be most accurately detected with sensors on the back and using a combination of orientation and angular velocity. The same setup also gives the best prediction of turn direction and amplitude. Preliminary measurements with a single amputee were also performed and highlighted important differences such as slower turning that need to be taken into account.

Previous studies have presented algorithms for detection of turns during gait using wearable sensors, but those algorithms were not built for real-time use. This paper therefore investigates the optimal approach for real-time detection of planned turns during gait using wearable inertial measurement units. Several different sensor positions (head, back and legs) and three different detection criteria (orientation, angular velocity and both) are compared with regard to their ability to correctly detect turn onset. Furthermore, the different sensor positions are compared with regard to their ability to predict the turn direction and amplitude. The evaluation was performed on ten healthy subjects who performed left/right turns at three amplitudes (22, 45 and 90 degrees). Results showed that turn onset can be most accurately detected with sensors on the back and using a combination of orientation and angular velocity. The same setup also gives the best prediction of turn direction and amplitude. Preliminary measurements with a single amputee were also performed and highlighted important differences such as slower turning that need to be taken into account.

This paper presents an approach for early recognition of gait initiation and termination using wearable inertial measurement units and pressuresensitive insoles. Body joint angles, ground reaction force and center of plantar pressure of each foot are obtained from these sensors and input into a supervised learning algorithm. For gait initiation, the algorithm detects two events: gait onset (the first detectable change from the baseline state) and toe-off. For gait termination, the algorithm segments gait into different steps, measures the signals over a window at the beginning of each step, and determines whether the measurement belongs to the final step. The approach is validated with 10 subjects at two different gait speeds, with both within-subject and subject-independent crossvalidation. Results show that the inertial measurement units are generally more useful than insoles during both gait initiation and termination, though combining both types of sensors results in better onset detection and easier segmentation of gait into different steps. However, for best performance the algorithms should be trained for each subject separately, and the gait termination recognition algorithm is not very robust with regard to gait speed.

This paper presents algorithms for detection of gait initiation and termination using wearable inertial measurement units and pressure-sensitive insoles. Body joint angles, joint angular velocities, ground reaction force and center of plantar pressure of each foot are obtained from these sensors and input into supervised machine learning algorithms. The proposed initiation detection method recognizes two events: gait onset (an anticipatory movement preceding foot lifting) and toe-off. The termination detection algorithm segments gait into steps, measures the signals over a buffer at the beginning of each step, and determines whether this measurement belongs to the final step. The approach is validated with 10 subjects at two gait speeds, using within-subject and subject-independent cross-validation. Results show that gait initiation can be detected timely and accurately, with few errors in the case of within-subject cross-validation and overall good performance in subject-independent cross-validation. Gait termination can be predicted in over 80% of trials well before the subject comes to a complete stop. Results also show that the two sensor types are equivalent in predicting gait initiation while inertial measurement units are generally superior in predicting gait termination. Potential use of the algorithms is foreseen primarily with assistive devices such as prostheses and exoskeletons.

The paper presents a wearable sensory system aimed at tracking of motion parameters and estimation of kinematic data of the wearer for use in controlling active lower-limb ortho-prostheses. The sensory system comprises inertial and magnetic measurement units (IMUs) attached to body segments and sensorized insoles worn inside sneaker shoes. Through sensory fusion, the IMU data is used to produce estimates of segment orientations, while the insoles provide information on ground reaction force amplitude and distribution. The paper outlines the principles of the algorithms and show the evaluation of them. Algorithms use both reference and wearable sensory data to extract information about the subject&#39;s kinematics, movement type and phase, and track selected biomechanical stability descriptors. The paper discusses preliminary experimental results of the proposed algorithms.

This paper presents a study of gait intention detection using force plates, inertial measurement units and an optical measurement system. The main goal is to detect gait initiation before heel-off and toe-off. Several established supervised machine learning methods are used to detect the onset of gait initiation, the first heel-off and the first toe-off. Events manually annotated by an expert serve as a reference. Results show that force plate signals are the most useful sensor, allowing gait onset to be detected with a mean absolute error of 0.12 seconds. Inertial measurement units are less accurate, with a mean absolute error for gait onset detection of 0.22 seconds. However, the decreased accuracy is primarily due to a small number of poorly detected outliers. The accuracy of the different supervised methods is also compared. For practical use, we recommend a combination of inertial measurement units and in-shoe pressure sensors, with different supervised methods used to detect different events.