LAUSR

Abstract This paper presents a real-time hierarchical approach to sound signal classification for utilization in hearing improvement devices. The developed classification hierarchy consists of three levels to classify speech, music and different noise types. A distinguishing attribute of this hierarchical approach is that effective features are computed as needed at different levels of the hierarchy making the classification process computationally efficient. This approach is compared to the conventional one-step classification approach by examining both trained and non-trained sound signals. The results obtained show higher classification rates as well as higher computational efficiency of this hierarchical approach compared to the conventional one-step approach.