An array of physical sensors and an adaptive regression strategy for emotion recognition in a noisy scenario

Abstract

30Several studies demonstrate that since emotions are spontaneously manifested through different 31 measurable quantities (e.g. vocal and facial expressions), this makes possible a sort of automatic estimation 32 of emotion from objective measurements. However, the reliability of such estimations is strongly influenced 33 by the availability of the different sensor modalities used to monitor the affective status of a subject, and 34 furthermore the extraction of objective parameters is sometime thwarted in a noisy and disturbed 35 environment. This paper introduces a personalized emotion estimation based on a heterogeneous array of 36 physical sensors for the measurement of vocal, facial, and physiological (electro-cardiogram and electro-37 dermal) activities. As a proof of concept, changes in the levels of both emotion reactiveness and pleasantness 38are estimated under critical operative conditions. The estimator model takes advantage from the time-39varying selection of the most relevant non-spurious sensors features and the adaptation of the k-nearest 40 neighbour paradigm to the continuous identification of the most affine model templates. The model, once 41trained, demonstrated to autonomously embed new sensorial input and adapt to unwanted/unpredicted 42 sensor noise or emotion alteration. The proposed approach has been successfully tested on the RECOLA 43 database, a multi-sensorial corpus of spontaneous emotional interactions in French. 44

An emotional modulation model as signature for the identification of children developmental disorders

“…The DD model is indeed constructed using a supervised learning approach ran over different individuals, thus requiring a cross-validation procedure for training and validation. In particular, by implementing a leave-one-patient-out cross validation procedure, we ran the dynamic feature selection (DFS) approach (see Section 3.3) to dynamically select the features according to each test data 37 , 38 . A binary classification model based on an SVM, with linear kernel and default parameters setting, is then trained on the EMF matrix of all the subjects except for one that, in turn, is left out for test.…”

“…Under the assumption that features selected play a crucial role in the recognition performance, especially due to the heterogeneity of the test set, we applied here a dynamic feature selection (DFS) procedure intended to optimally select model features according to each specific test data. More specifically, in line with the recently developed methodology 37 , 38 , we design the following three-level DFS approach:…”

Online Feature Selection for Robust Classification of the Microbiological Quality of Traditional Vanilla Cream by Means of Multispectral Imaging

“…An important point in this work is that the developed model included as much variability as possible, as in real life, since model development was based on two different product batches supplied directly by the manufacturer of the vanilla cream that were maintained under laboratory conditions, whereas model testing and validation were based on samples from different retail outlets as well as expired samples from the market. The performance of UOS approach has been successfully used in the past with simulated and experimental datasets for robust classification with drifting and faulty gas sensors [26] as well as in the case of self-repairing classification algorithms for chemical sensor array [55], affective computing [33], and developmental disorders recognition [34]. Not by chance all the considered scenarios present an intrinsic data heterogeneity that makes it usually difficult to develop an effective discrimination strategy using standard paradigms (i.e., diversity of emotion manifesting, autism phenotyping, food spoilage distribution).…”

“…Under the assumption that feature selection plays a crucial role in the recognition performance, especially due to the heterogeneity of the test set, a UOS based on a dynamic feature selection (DFS) procedure intended to optimally select model features according to each specific test data was employed in this work. More specifically, in line with the UOS approach recently developed [26,33,34] the following three-level DFS approach was designed, as summarized in Figure 2.…”

Deciphering Cancer Cell Behavior From Motility and Shape Features: Peer Prediction and Dynamic Selection to Support Cancer Diagnosis and Therapy

“…As an example, in a group of cells moving toward a target cell, e.g., immunecancer cross-talk (13,14), speed and directional persistence are needed to model their collective motion; on the other hand, in a group of cells interacting with a target cell, e.g., immune cells killing a cancer cell (15,16), mean interaction time and track curvature have proved to be specifically tailored for the phenomenon quantification. In particular, in this work, we extended and applied a dynamic feature selection (DFS) procedure (17,18), selecting, in an unsupervised way, the optimal feature set extracted from the training set for each new test sample; this will be used to build a classifier for the test label prediction.…”