We developed a music information retrieval system based on singing voice timbre, i.e., a system that can search for songs in a database that have similar vocal timbres. To achieve this, we developed a method for extracting feature vectors that represent characteristics of singing voices and calculating the vocal-timbre similarity between two songs by using a mutual information content of their ...

In this paper a humming method for music information retrieval is presented. The system uses a database with real songs and does not need another type of symbolic representation of them. The system employs an original fingerprint based on chroma vectors to characterize the humming and the references songs. With this fingerprint, it is possible to get the hummed songs without needed of transcrip...

This document describes our submission to QBSH task of MIREX 2013. Our algorithm adopts a two-stage cascaded solution based on Locality Sensitive Hashing (LSH) and accurate matching of frame-level pitch sequence. Firstly, LSH is employed to quickly filter out songs with low matching possibilities, resulting in a list of candidate songs for further processing. In the second stage, Dynamic Time W...

In hearing insects, discrimination of auditory signals is vital to mate selection. We tested the capability of single auditory receptor neurons to discriminate calling songs from di!erent individual grasshoppers of the same species. Spike trains elicited from di!erent songs were studied using the metric-space analysis of Victor and Purpura (Networks: Comput. Neural Syst. 8 (1997) 127}164). Our ...

Searching and organizing growing digital music collections requires automatic classification of music. Our system for artist and genre identification uses support vector machines to classify songs based on features calculated over their entire lengths. Since support vector machines are exemplar-based classifiers, training on and classifying entire songs instead of short-time features makes intu...