Abstarct: In recent years, significant researches has been done in the field of environmental sounds and audio events recognition, but these studies are very limited in comparison with other related areas such as speech and music. The aim of this thesis is the development of feature extraction methods for sound events recognition in office environment. The IEEE Audio and Acoustic Signal Processing Technical Committee Challenge on Detection and Classification of Acoustic Scenes and Events (DCASE) dataset is used for the task. The dataset consists of 16 classes of sound events in office environment, which some of these sounds are noisy with different SNR. For this research, two feature extraction methods are introduced. In the first method, features are extracted by Matching Pursuit (MP) algorithm in combination with common Mel Frequency Cepstral Coefficients (MFCC) as feature vectors by using nearest neighbour classifier. The recognition rate is 69.67 percent in this method that is increaded 6 percent in comparison with the case of using MFCC without MP. In the second method, Bark Frequency Cepstral Coefficients (BFCC) and GMM are employed as feature and classifier, respectively. In this case, recognition rate is achieved 80.08 percent that shows effectiveness of proposed method in comparison with the most of existing methods for used dataset.