Abstract
In this research, an algorithm was proposed to automatically classify the mood of the speaker by referring to his speech. Three moods were adopted in this study, namely joy, sadness and anger in order to distinguish between them. The principle of the algorithm work includes the initial treatment of the signal of by removing the silence and then cut the signal to a number of sections length of each 512 sample, and then treatment by window (Hamming window) followed by the process of extracting the characteristics such as energy, the basic frequency, resonance frequencies of each section and for all speech signals that Were recorded, which included 30 signals of persons between 15 and 25 years of age in order to prepare the database for the three moods and to draw the characteristic curves and for each mood. The selection of signals was done from training and testing set for detectingthe mood of these signals by performing the previous steps and then comparing the resulting curve with the previous curves using the correlation coefficient and the Euclidean distance. The algorithm gave good results when these characteristics were adopted in the classification process and by about 75%.
