INFORMATION AND COMMUNICATION TECHNOLOGIES AND SERVICES VOLUME: 10 | NUMBER: 4 | 2012 | SPECIAL ISSUE 270 © 2012 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING FUNDAMENTAL FREQUENCY EXTRACTION METHOD USING CENTRAL CLIPPING AND ITS IMPORTANCE FOR THE CLASSIFICATION OF EMOTIONAL STATE Pavol PARTILA 1 , Miroslav VOZNAK 1 , Martin MIKULEC 1 , Jaroslav ZDRALEK 1 1 Department of Telecommunications, Faculty of Electrical Engeneering and Computer Science, VSB–Technical University of Ostrava, 17. Listopadu 15, 708 33 Ostrava-Poruba, Czech Republic pavol.partila@vsb.cz, miroslav.voznak@vsb.cz, martin.mikulec@vsb.cz, jaroslav.zdralek@vsb.cz Abstract. The paper deals with a classification of emotional state. We implemented a method for extracting the fundamental speech signal frequency by means of a central clipping and examined a correlation between emotional state and fundamental speech frequency. For this purpose, we applied an approach of exploratory data analysis. The ANOVA (Analysis of variance) test confirmed that a modification in the speaker's emotional state changes the fundamental frequency of human vocal tract. The main contribution of the paper lies in investigation, of central clipping method by the ANOVA. Keywords Central clipping, DC offset, emotional state, features extraction, hamming smoothing window, homoscedasticity, pre-emphasis. 1. Introduction Man-machine interaction is a desirable trend, accompanied hand in hand with an effort to improve the quality of mutual communication. On the other hand, we feel the absence of credibility of information presented by a synthetic speech from a computer’s loudspeaker. Speeches generated by Text-to-Speech tools act artificially because they do not take into account the emotional state. In speech, the emotional state is characterized by specific phonetic features. These features include intensity, intonation and timbre of speech. In the domain of speech processing, the speech signals are described by parameters such as signal energy, zero crossing ratio and fundamental speech frequency or by cepstral coefficients [1], [2], [3]. 2. Pre-processing Once human speech is digitalized, the digital audio record can be analyzed. In order to extract signatures such as the fundamental speech signal frequency, energy, etc., it is necessary to carry out several operations depicted in Fig. 1. These steps need to be carried out before the above-mentioned signatures have been extracted [4]. DC Offset Preemphase Segmentation Windowing Fig. 1: Pre-processing of speech signals. 2.1. DC Offset A number of audio cards add DC (Direct Current) components into the audio signal, as depicted in Fig. 2. Approaches used in digital signal processing are applied to compute some signatures. The DC component in the signal negatively affects the computation and may cause disturbance. Fig. 2: Effect of DC Offset on speech signal. It is therefore necessary to remove the DC component before the processing. The DC component of