Етикет: интелигентни системи

Key Words: Deep Neural Networks; Facial Emotions Recognition; Weather Condition Recognition; Python.

Abstract. Emotions are one of the main ways to communicate between people and to express their attitude towards objects, products, services, etc. Emotions are divided into two classes – verbal and non-verbal. Human speech and intonation belong to the first class, and to the second class are facial and body emotions, also known as body language. The subject of this paper is facial emotions and their relationship to the scene in which they occur. A number of studies have established that there is a strong relationship between a person’s emotions and their surroundings. The latter includes meteorological conditions (weather) and other objects, such as other people, landscape, etc. Facial emotions range (FER) from seven basic emotions (joy, anger, surprise, fear, sadness, neutral and disgust and neutral) categorized by P. Ekman through his Facial Action Coding System to 26 emotions represented by Russell through his 3D Valence Arousal Dominance model. Most of the existing deep neural networks for Facial Emotions Recognition recognize mentioned seven emotions. In our previous research, we presented a pre-trained FER model with 69.85% accuracy. Weather conditions are closely related to geographic regions and vary in some cases from sunny to cloudy, or in other cases include some subset of sunny, foggy, snowy, rainy, hot, etc. In this research, we analyze deep learning neural networks, for weather conditions recognition and selected appropriate model. We combined our FER DNN with the selected weather recognition DNN and build a bimodal system, which improves facial emotion recognition to 80-83% especially in the cases when FER model provides contradictory results.

Повече

Key Words: Online Learning; Deep Neural Networks; Face Recognition; Facial Emotions Recognition; Python.

Abstract. Current study presents development of bimodal system for Facial Emotion Recognition (FER) and Body Gestures Emotion Recognition (BER). The system is based on two Deep Learning Neural Networks (DNN) each one responsible for the recognition of the emotion of the face or the body. The use of the combination of two neural networks has an amplifying synergistic effect, which increases by about 10% the accuracy of the results (recognized emotionс) compared to those of the individual DNN. The selection of pre-trained DNN models for facial and body emotions recognition is based on two authors’ papers, in which detailed analysis of the DNN for FER and BER has been done. Therefore in current study a brief information about selected DNN models is provided, as well information about specific dataset used for training selected DNNs. Verification of the bimodal system is done using our private dataset.

Повече

Key Words: Waste collection; smart city; Internet of Things (IoT); intelligent transportation systems; surveillance systems.

Abstract. The paper presents the structure of an intelligent integrated system for managing the collection of solid waste in urban and suburban environments. The system can automatically maintain the box level and send information to the waste collection truck. The technologies used in the proposed system are good needed to provide real monitoring and management of waste collection processes and to obtain a green environment.

Повече

Key Words: NIR spectroscopy; Partial Least Square method; Radial Basis Functions Neural Networks; moisture content; fat content; pork meat samples.

Abstract. In this study the performance of linear and nonlinear chemometric methods has been investigated and compared. The transmittance spectra of pork meat samples were collected by SW-NIR (short wave near infrared) analyser in the spectral range of 850 nm to 1,050 nm. Partial Least Square (PLS1) method and Radial Basis Functions Neural Networks (RBF NN) were chosen for chemometric analysis of that samples for determination of moisture and fat content. The reason for using RBF ANN is significant nonlinearity which is exhibited between the spectra and the fat and moisture content. PLS1 and RBF NN with different architecture have been combined with different pre-processing techniques such as first derivative (D1), standard normal variate (SNV), multiplicative signal correction (MSC) and the combinations of MSC and SNV with first derivative. It was found that optimal pre-processing was MSC for moisture, and the combination of D1and SNV for fat. When PLS1 was used, results showed reduction of RMSEP and REP using MSC with 15 and 13% for moisture determination. In case of PLS1 fat determination considerable reduction of RMSEP and REP was observed using a combination of D1 and SNV with 48 and 47%. Compared to PLS1 regression with suitable preprocessing, RBF ANN showed better results: reduction of RMSEP and REP using a combination of D1 and SNV with 48% for moisture, and reduction using a combination of D1 and SNV with 59% for fat determination. These improvements together with the facility of SW-NIR technology to be implemented in the process engineering made it ideal for the meat industry.

Повече

Key Words: Internet of Things; Wireless Sensor Network; smart home; integration; voice assistant.

Abstract. An integrated Smart Home system for monitoring and management of the elements of the working environment or at home in Home Assistant platform and is integrated with a voice assistant (Google Assistant) has been developed. The structure of the Smart Home (SH) system is based on the concept of the Internet of Things (IoT), which includes connectivity of devices and actuators, as well as the presence of Wireless Sensor Network (WSN). The main functions in each SH depend strongly on the requirements, way of live, special health issues, availability of pets and appliances etc. of the household members, as well as the home/house architecture, location etc. The developed integrated system structure consists of the following main modules: connected devices, measuring (sensor) module, processing module, visualization module (interface), communication module (voice communication). The system allows monitoring and control of various parameters of the environment, determination of geolocation, tracking the state of the connected devices, provides ascertainment of conditions or constraints during the implementation of logical algorithms or actions, etc. The developed integrated system solves the problem of using various interface applications, communication protocols and standards by integration of all its elements in one Application Programming Interface (API) and simultaneously the system is expanding its scope through its integration with a voice assistant (Google Assistant). In the developed integrated system solutions with pre-set functions, default functions and user selection functions are implemented. Also, specially designed by the author (G. Kolev), made and tested IoT boards for stepper motor control, RGB LED strip, as well as IoT board for control of small relays for concealed mounting are applied. The system operation rules can be set by the user (directly or in time) or depending on the values obtained from the sensors (Sensor-based Linked Open Rules). The developed SH system gives also the possibility for building actions according to the geolocation of each of the devices (users) via the GPS system of the phone. Elements, connected with the energy utilization (consumption) efficiency – electricity, water and heat consumption efficiency are also considered. The optimization of the consumption is directly connected with cost savings which adds an additional benefit to the undeniable advantages of the Smart Home system development. The process of development of integrated remote-control Smart Home systems can continue without limitation in time as far as the imagination of the designer and/or the users reaches.

Повече

Key Words: Collaborative work; humans and robots; standards; risk management.

Abstract. Robots are a huge part of all human activities. In the industry is proven that collaborative work of humans and robots is more effective, than the production technologies without humans. This, in turn, a new range of problems was created, both in terms of productivity and in safe collaboration, especially for people. The present work sets out to solve the most essential problems for ensuring conflict-free operation while guaranteeing high performance and quality of the functional tasks performed.

Повече

Key Words: Deep Neural Networks; R-CNN; object detection; classification; safety system.

Abstract. Current paper presents an application of Deep Neural Networks in the field of detecting dangerous objects and violence in different areas. Using machine vision, combined with Deep Neural Networks makes possible to improve the safety of citizens in different areas such as the mall, local store and others where guards cannot act fast enough in case of a danger situation.

Повече

Key Words: Content recommendation; gamе-based learning; special educational needs; personalized learning; machine learning.

Abstract. This article presents an approach for creating a content recommendation model in a gamе-based learning platform designed for students with special educational needs. Various referral systems and existing opportunities for intelligent delivery of personalized learning content tailored to the individual needs of children are addressed.

Повече