Angélica Muñoz-Meléndez

Classification of bipolar disorder episodes based on analysis of voice and motor activity of patients

There is growing amount of scientific evidence that motor activity is the most consistent indicator of bipolar disorder. Motor activity includes several areas such as body movement, motor response time, level of psychomotor activity, and speech related motor activity. Studies of motor activity in bipolar disorder have typically used self-reported questionnaires with clinical observer-rated scales, which are therefore subjective and have often limited effectiveness. Motor activity information can be used to classify episode type in bipolar patients, which is highly relevant, since severe depression and manic states can result in mortality. This paper introduces a system able to classify the state of patients suffering from bipolar disorder using sensed information from smartphones. We collected audio, accelerometer and self-assessment data from five patients over a time-period of 12 weeks during their real-life activities. In this research we evaluated the performance of several classifiers, different sets of features and the role of the questionnaires for classifying bipolar disorder episodes. In particular, we have shown that it is possible to classify with high confidence ( ź 85 % ) the course of mood episodes or relapse in bipolar patients. To our knowledge, no research to date has focused on naturalistic observation of day-to-day phone conversation to classify impaired life functioning in individuals with bipolar disorder.

Estimation of temporal gait parameters using Bayesian models on acceleration signals

The purpose of this study is to develop a system capable of performing calculation of temporal gait parameters using two low-cost wireless accelerometers and artificial intelligence-based techniques as part of a larger research project for conducting human gait analysis. Ten healthy subjects of different ages participated in this study and performed controlled walking tests. Two wireless accelerometers were placed on their ankles. Raw acceleration signals were processed in order to obtain gait patterns from characteristic peaks related to steps. A Bayesian model was implemented to classify the characteristic peaks into steps or nonsteps. The acceleration signals were segmented based on gait events, such as heel strike and toe-off, of actual steps. Temporal gait parameters, such as cadence, ambulation time, step time, gait cycle time, stance and swing phase time, simple and double support time, were estimated from segmented acceleration signals. Gait data-sets were divided into two groups of ages to test Bayesian models in order to classify the characteristic peaks. The mean error obtained from calculating the temporal gait parameters was 4.6%. Bayesian models are useful techniques that can be applied to classification of gait data of subjects at different ages with promising results

People Detection by a Mobile Robot Using Stereo Vision in Dynamic Indoor Environments

People detection and tracking is a key issue for social robot design and effective human robot interaction. This paper addresses the problem of detecting people with a mobile robot using a stereo camera. People detection using mobile robots is a difficult task because in real world scenarios it is common to find: unpredictable motion of people, dynamic environments, and different degrees of human body occlusion. Additionally, we cannot expect people to cooperate with the robot to perform its task. In our people detection method, first, an object segmentation method that uses the distance information provided by a stereo camera is used to separate people from the background. The segmentation method proposed in this work takes into account human body proportions to segment people and provides a first estimation of people location. After segmentation, an adaptive contour people model based on people distance to the robot is used to calculate a probability of detecting people. Finally, people are detected merging the probabilities of the contour people model and by evaluating evidence over time by applying a Bayesian scheme. We present experiments on detection of standing and sitting people, as well as people in frontal and side view with a mobile robot in real world scenarios.

A minority class feature selection method

In many classification problems, and in particular in medical domains, it is common to have an unbalanced class distribution. This pose problems to classifiers as they tend to perform poorly in the minority class which is often the class of interest. One commonly used strategy that to improve the classification performance is to select a subset of relevant features. Feature selection algorithms, however, have not been designed to favour the classification performance of the minority class. In this paper, we present a novel filter feature selection algorithm, called FSMC, for unbalanced data sets. FSMC selects attributes that have minority class distributions significantly different from the majority class distributions. FSMC is fast, simple, selects a small number of features and outperforms in most cases other feature selection algorithms in terms of global accuracy and in terms of performance measures for the minority class such as precision, recall, F-measure and ROC values.

Collaborative robots for indoor environment exploration

The design and implementation of a robot team consisting of three homogeneous mobile robots and an external server is presented. The main goal of the robot team is the updating of a map or representation of an indoor environment. A scheme for collective exploration was defined. This scheme enables robots to navigate and self-locate within an indoor environment, communicate to each other, create local maps of their environment to be merged into a global map, and coordinate individual actions in order to explore autonomously the environment. Finally, the performance of the robot team was evaluated in various environmental conditions.

Using Intermediate Models and Knowledge Learning to Improve Stress Prediction

Motor activity in physical and psychological stress exposure has been studied almost exclusively with self-assessment questionnaires and from reports that derive from human observer, such as verbal rating and simple descriptive scales. However, these methods are limited in objectively quantifying typical behaviour of stress. We propose to use accelerometer data from smartphones to objectively quantify stress levels. Used data was collected in real-world setting, from 29 employees in two different organisations over 5 weeks. To improve classification performance we propose to use intermediate models. These intermediate models represent the mood state of a person which is used to build the final stress prediction model. In particular, we obtained an accuracy of 78.2 % to classify stress levels.

Learning Concepts with Multi-Robot Systems

This paper introduces a novel approach to learn representations of objects using a team of robots. Each robot extracts local and global visual features of objects and combines them to represent and recognize objects. Contrary to previous approaches the robots do not know in advance the number or nature of objects to learn. Individual representations of objects are learned on-line while the robots are traversing an environment. Robots share their individual concepts to improve their own concepts, and to acquire a new representation of an object not seen by them. For that, the robots have to detect if they are seeing a new object or an already learned one. We empirically evaluated our approach with a real world robot team with very promising results.

Detection of multiple people by a mobile robot in dynamic indoor environments

Detection of multiple people is a key element for social robot design and it is a requirement for effective human-robot interaction. However, it is not an easy task, especially in complex real world scenarios that commonly involve unpredictable motion of people. This paper focuses on detecting multiple people with a mobile robot by fusing information from different sensors over time. The proposed approach applies a segmentation method that uses the distance to the objects to separate possible people from the background and a novel adaptive contour people model to obtain a probability of detecting people. A probabilistic skin model is also applied to the images and both evidences are merged and used over time with a Bayesian scheme to detect people. We present experimental results that demonstrate how the proposed method is able to detect people who is standing, sitting and leaning sideways using a mobile robot in cluttered real world scenarios.

Simple linear vision module for micro mobile robot applications

A low-cost electronic device that transforms patterns of light into a digital signal recognized by a common electronic board to control micro mobile robots is presented. The device has been tested with a micro mobile robot for recognizing and reading bar codes within a structured environment.

ADAPTIVE LOCOMOTION FOR A HEXAGONAL HEXAPOD ROBOT BASED ON A HIERARCHICAL MARKOV DECISION PROCESS

This article describes a probabilistic control model based on Markov Decision Processes for the locomotion of a hexagonal hexapod robot. Uncertainty is naturally taken into account in probabilistic models, resulting in flexible control models that enable a robot to react to both, expected and unexpected situations. The model was tested using a simulated robot under various experimental conditions.