The amount of audio data on public networks like Internet is increasing in huge volume daily. So to access these media, we need to efficiently index and annotate them. Due to non-stationary nature and discontinuities present in the audio signal, segmentation and classification of audio signal has really become a challenging task. Automatic music classification and annotation is also one of the challenging tasks due to the difficulty in extracting and selecting optimal audio features. Today, content-based audio retrieval systems are used in various application domains and scenarios such as music retrieval, speech recognition, and acoustic surveillance. During the development of an audio retrieval system, a major challenge is the identification of appropriate content-based features for representation of the audio signals under consideration. This paper gives the overview of various techniques used for classification and retrieval of audio and also proposes a novel approach for classification and retrieval of audio signal.