It has been established that it is possible to reveal human emotions using electroencephalogram (EEG) signals. Most studies used a wide variety of data sets and methods, therefore a comparison between the performances of their approaches is difficult. This paper reports a study on the effects of the number of electrode channels and frequency bands for emotion classification based on a database for emotion analysis using physiological signals (DEAP). Discrete wavelet transform (DWT) was used for feature extraction and support vector machine (SVM) was applied as the classifier. From experimental results, it is found that (a) using more electrodes channels did not guarantee better accuracy, (b) comparing with all frequency bands, using a few of them did not reduce the accuracy dramatically and some results revealed that only two bands produced better results, and, (c) the more emotions to be classified, the lower accuracy was achieved based on the same method.