Similarity-based visualization of sounds

The thesis focuses on the visualization of a collection of sound recordings by the criterion of mutual similarity. The product of the thesis is an online application, which includes a graphic visualization of the calculated data and metadata of the observed recordings. The studied group consists of sound recordings, which we acquired from the website of the Work With Sounds project and a few other sites. First we calculated 27 low-level features of each recording, which were the basis on which we determined the similarities. Then we used hierarchical clustering to combine the recordings into groups. The final visualization was made with SVG graphic elements, which are based on data and generated by the D3 JavaScript library. We used the library to realise the user interaction and make it possible to see the picture of sound and play the recording. The visualization is based on the zoomable circle packing method