With the rapid increase in the capability of data storage, it has become more and more important to develop automated techniques for data classification. In particular, multimedia material like audio is difficult to browse, because it takes time and energy to play and listen to the audio data. Recognizing this, our work tries to design a system for verifying whether or not an unknown file in a personal computer (PC) belongs to audio data and further identifying which acoustic class the file is, if the file belongs to audio. The acoustic classes we considered here encompass noise, speech, singing, instrument-sole music, and accompanied singing. For the case that a test file is speech, singing, or accompanied singing, we further identify the gender of the person involved in the audio data, e.g., male/female speaker or male/female singer. The proposed system is of a hierarchical structure that divides audio files into nine classes. Our experiments show that the system is feasible in PC file management.