polito.it
Politecnico di Torino (logo)

Automatic Recognition And Classification Of Passengers' Emotions In Autonomous Driving Vehicles

Antonio Costantino Marceddu

Automatic Recognition And Classification Of Passengers' Emotions In Autonomous Driving Vehicles.

Rel. Massimo Violante, Jacopo Sini. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Informatica (Computer Engineering), 2019

[img]
Preview
PDF (Tesi_di_laurea) - Tesi
Licenza: Creative Commons Attribution.

Download (20MB) | Preview
Abstract:

Automatic Recognition And Classification Of Passengers' Emotions In Autonomous Driving Vehicles Introduction Development of autonomous driving vehicles is a challenging activity, with technological, ethical, and social implications. People use cars for over a hundred years, and during all this time period the vehicles have been operated by human drivers. Due to this habit, it is simpler for people to trust another person with respect to a computer to drive their vehicles. So, it is important to consider their social implications in order to avoid making the expected safety improvement useless due to a lack of trust of people on those vehicles. These regards ethic issues, social trust on autonomous driving capabilities, and novel commercialization model for car manufacturer. So, new methods for helping the social acceptance and trustiness on these new technologies are needed. Thesis purpose Our idea to solve this problem, that is the main purpose of this thesis, is to perform autonomous vehicles passengers’ emotions detection and use this data to adapt the autonomous cars driving style. Regarding the driving style adaptation, we’ll make some clarifications about what is the main objective that we want to achieve: - if the people on the car are scared or sad, the car adopts a caring driving style, that slows the speed and takes the curve more accurately (lowering the lateral accelerations); - if the people on the car are neutral, the car adopts a normal driving style; - if the people on the car are happy or over enjoyed, the car adopts a sportive driving style (steeper acceleration/braking ramps and curve trajectory with higher lateral accelerations). In this way, we hope to improve people’s confidence towards self-driving cars. Facial Expression Database Classifier As first step, we focused on the dataset preparation activities. To improve the effectiveness of this phase, we chose to develop a novel software that we called Facial Expression Database Classifier (FEDC); FEDC is a program able to automatically classify images of some of the most used databases, depicting posed human faces: - Extended Cohn-Kanade Database (CK+); - FACES Database; - Facial Expression Recognition 2013 Database; - Japanese Female Facial Expression Database (JAFFE); - Multimedia Understanding Group Database (MUG); - Radboud Faces Database (RaFD); - Static Facial Expression in the Wild 2.0 Database (SFEW 2.0). Neural Network After realizing the first version of FEDC, we started to work on the possible improvements for this program and we also started to work on the implementation of a neural network. After a period of research of the state of the art regarding neural networks capable of doing emotion detection, we decided to implement the neural network proposed in the “Physiological Inspired Deep Neural Networks for Emotion Recognition” paper , obtaining almost the same results as the authors. Road Tests The last part of our work was to make some road test, in order to test the neural network performance. So, we developed Emotion Detector, a program that can take picture from the camera of a laptop and predict the emotions of a person. Using this tool, we took some pictures on realistic condition, obtaining very good results. Conclusion After the test, and in summary, we think that the result we obtained are really promising: this approach could really facilitate the diffusion of the autonomous driving cars.

Relatori: Massimo Violante, Jacopo Sini
Anno accademico: 2019/20
Tipo di pubblicazione: Elettronica
Numero di pagine: 100
Soggetti:
Corso di laurea: Corso di laurea magistrale in Ingegneria Informatica (Computer Engineering)
Classe di laurea: Nuovo ordinamento > Laurea magistrale > LM-32 - INGEGNERIA INFORMATICA
Aziende collaboratrici: NON SPECIFICATO
URI: http://webthesis.biblio.polito.it/id/eprint/12423
Modifica (riservato agli operatori) Modifica (riservato agli operatori)