Affective computing studies and develops systems capable of detecting humans affects. The search for universal well-performing features for speech-based emotion recognition is ongoing. In this paper, a small set of features with support vector machines as the classifier is evaluated on Surrey Audio-Visual Expressed Emotion database, Berlin Database of Emotional Speech, Polish Emotional Speech database and Serbian emotional speech database. It is shown that a set of 87 features can offer results on-par with state-of-the-art, yielding 80.21, 88.6, 75.42 and 93.41% average emotion recognition rate, respectively. In addition, an experiment is conducted to explore the significance of gender in emotion recognition using random forests. Two models, trained on the first and second database, respectively, and four speakers were used to determine the effects. It is seen that the feature set used in this work performs well for both male and female speakers, yielding approximately 27% average emotion recognition in both models. In addition, the emotions for female speakers were recognized 18% of the time in the first model and 29% in the second. A similar effect is seen with male speakers: the first model yields 36%, the second 28% a verage emotion recognition rate. This illustrates the relationship between the constitution of training data and emotion recognition accuracy.
The heat supply systems energy efficiency improvement requires the use of increasingly complex methods. The basic ways to reduce heat consumption is by using better thermal insulation, although they have more and more limited possibilities and need relatively large financial outlays. Good effects can be achieved by the better heat source adaptation to the conditions of a specific facility supplied with heat. However, this requires research that identifies the effectiveness of such solutions as well as the tools used to describe selected elements of the system or its entirety. The article presents the results of tests carried out for a gas boiler room supplying heat to a group of residential buildings. The goal was to build a model that would forecast the day range in which the maximum gas consumption occurs for a given day. Having measurements of gas consumption in subsequent hours of the day, it was decided to build a forecasting model determining the part of the day in which such a maximum would occur. To create the model the random forest procedure was used along with the mlr (Kassambara) package. The model’s hyperparameters were tuned based on historical data. Based on data for another period of boiler room operation, the results of the model’s quality assessment were presented. Close to 44% efficiency was achieved. Tuning the model improved its predictive ability.
In this paper, the future Fifth Generation (5G New Radio) radio communication system has been considered, coexisting and sharing the spectrum with the incumbent Fourth Generation (4G) Long-Term Evolution (LTE) system. The 4G signal presence is detected in order to allow for opportunistic and dynamic spectrum access of 5G users. This detection is based on known sensing methods, such as energy detection, however, it uses machine learning in the domains of space, time and frequency for sensing quality improvement. Simulation results for the considered methods: k-Nearest Neighbors and Random Forest show that these methods significantly improves the detection probability.
en
pl
