Feature Space Modification for Content-Based Music Retrieval Based on User Preferences

Abstract

This paper proposes a feature space modification method for feature extraction of music, which is effective for the development of a content-based music information retrieval (MIR) system based on user preferences. The proposed method conducts clustering of all songs in the music collection, and utilizes the resulting cluster IDs as training data for feature space modification, and is capable to automatically generate a feature space which is suitable to the content of any music collection. Experiment results prove that the proposed method improves accuracy of user preference based MIR.

Feature Analysis and Normalization Approach for Robust Content-Based Music Retrieval to Encoded Audio with Different Bit Rates

“…From the previous four sets of music audio files, we have constructed content-based MIR systems, following the method proposed in [2]. Namely, the feature space for each set (hereafter referred as: Raw_hist, 192_hist, 128_hist, and 64_hist, respective to the format/bit rate of the music collection) is generated by the method of [2], and all songs in each data collection are vectorized based on the corresponding feature space.…”

“…Namely, the feature space for each set (hereafter referred as: Raw_hist, 192_hist, 128_hist, and 64_hist, respective to the format/bit rate of the music collection) is generated by the method of [2], and all songs in each data collection are vectorized based on the corresponding feature space. Furthermore, in order to simulate a music collection composed of songs in various formats, we have also generated a "mixed" data collection of MP3 files, by randomly selecting the bit rate of each song evenly in the experimental data set.…”

“…For this experiment, we have developed a prototype MIR system, based on the MIR method proposed by Hoashi et al [2], which utilizes the tree-based vector quantization (TreeQ) algorithm proposed by Foote [3]. TreeQ constructs the feature space to vectorize music, based on training data, i.e., music with category labels.…”

Automatic Playlist Generation from Self-Organizing Music Map

“…In creating the playlist, one of the most important aspects of the method is the similarity measure used to compare two songs. We are aware that there are several methods for measuring the similarity of songs [13,14,15,16,17]. However, due to its simplicity and intuitiveness, we utilized the Euclidean distance to measure the similarity of the songs in our system.…”