Brahmsian articulation: Ambiguous and unfixed structures in Op. 38

Llorens, A. (2021). 38. Music Theory Online, 27(4).

Five different recordings of Brahms's first chello sonata, Op.38 are examined with respect to timing and dynamic patterns. Viscovery SOMine is used to cluster measures with regards to duration percentages of quarter notes and loudness changes between quarter notes.

Recorded asynchronies, structural dialogues: Brahms's Adagio Affettuoso, Op. 99ii, in the hands of Casals and Horszowski

Llorens, A. (2017). Music Performance Research, 8.

This article tries to find intended musical motives behind obvious asynchronies in Casals and Horszowski's 1935 performance of Brahms's Adagio Affettuoso, Op. 99ii. Clustering bars with respect to the timing of the semiquavers in Viscovery SOMine showed that there are very distinct and recurring asynchrony patterns, which are most likely not results of pure chance or local expressive devices, but rather critical elements of an ongoing structural dialogue between the musicians.

The form of performance: analyzing pattern distribution in select recordings of Chopin's Mazurka Op. 24 No. 2

Spiro, N., Gold, N., & Rink, J. (2010). Musicae Scientiae, 14(2), 23-55.

This study analyzes 29 performances of Chopin's Mazurka Op. 14 No. 2 to find recurring timing and dynamic patterns. Viscovery SOMine is used to cluster the four bar phrases in this music piece to better understand the variances and similarities between different performances from a single musician and from different musicians.

Recognition of Western style musical genres using machine learning techniques

Mostafa, M. M., & Billor, N. (2009). Expert Systems with Applications, 36(8), 11378-11389.

Several different machine learning techniques are used to classify different music genres with respect to frequency and amplitude characteristics. In addition to the classification model, Viscovery SOMine is used for visual exploration of genre differences.

Application methods for self-organizing map in process imaging for dynamic behavior of aerated agitation vessel

Matsumoto, H., Masumoto, R., & Kuroda, C. (2007). In Proceedings of the 10th International Conference on Engineering Applications of Neural Networks (pp. 210-220).

Viscovery SOMine is adapted to process imaging for dynamic behavior of an aerated agitation vessel, and various application methods are investigated. In the application, the direct imaging by CCD video camera and the PIV technology are adopted. The generated map and clusters could give process engineers useful information about the degree of spatial dispersion of bubbles as well as about the determination of design parameters.

Using self-organizing maps for object classification in Epo image analysis

Heiss-Czedik, D., & Bajla, I. (2005). Measurement Review, 5, 11-16.

The recombinant form of erythropoietin (rEpo), a hormone used for doping, can be detected via Epo chemiluminescence. A research project funded by the World Anti-Doping Agency has been established to develop software for Epo testing. For this purpose, 506 records from the training set are ordered by self-organizing maps using Viscovery SOMine, which was chosen primarily for its visualization capabilities. Moreover, Viscovery SOMine classification performs rather well on this problem when compared with other classification methods.

Neural networks for text-to-speech phoneme recognition

Embrechts, M. J., & Arciniegas, F. (2000). In Systems, Man, and Cybernetics, 2000 IEEE. International Conference on Systems, Man, and Cybernetics (Vol. 5, pp. 3582-3587). IEEE.

Two different artificial neural network (ANN) approaches are used for phoneme recognition for text-to-speech applications: staged backpropagation neural networks and self-organizing maps. Several current commercial approaches rely on an exhaustive dictionary approach for text-to-phoneme conversion. Applying neural networks to phoneme mapping for text-to-speech conversion creates a fast distributed recognition engine. This engine not only supports the mapping of missing words in the database, but it can also mitigate contradictions related to different pronunciations for the same word. The ANNs presented in this work were trained based on the 2,000 most common words in American English. Performance metrics for the 5,000, 7,000 and 10,000 most common words in English are also estimated to test the robustness of these neural networks.