Elżbieta Kubera

Musical Instruments in Random Forest

This paper describes automatic classification of predominant musical instrument in sound mixes, using random forests as classifiers. The description of sound parameterization applied and methodology of random forest classification are given in the paper. Additionally, the significance of sound parameters used as conditional attributes is investigated. The results show that almost all sound attributes are informative, and random forest technique yields much higher classification results than support vector machines, used in previous research on these data.

Identification of a dominating instrument in polytimbral same-pitch mixes using SVM classifiers with non-linear kernel

In this paper we deal with the problem of identification of the dominating musical instrument in a recording containing simultaneous sounds of the same pitch. Sustained harmonic sounds from one octave of twelve instruments were considered. The training data set contains isolated sounds of two forms, one from selected musical instruments, and the other from the same mixed with artificial harmonic and noise sounds of lower amplitude. The test data set contains mixes of musical instrument sounds. A Support Vector Machine classifier was used for training and testing experiments, using a non-linear kernel. Additionally, we performed tests on data based on different recordings of instruments than those used in the training procedure described above. Results of these experiments are presented and discussed.

Analysis of Recognition of a Musical Instrument in Sound Mixes Using Support Vector Machines

Experiments with recognition of the dominating musical instrument in sound mixes are interesting from the point of view of music information retrieval, but this task can be very difficult if the mixed sounds are of the same pitch. In this paper, we analyse experiments on recognition of the dominating instrument in mixes of same-pitch sounds of definite pitch. Sound from one octave (no. 4 in MIDI notation) have been chosen, and instruments of various types, including percussive instruments were investigated. Support vector machines were used in our experiments, and statistical analysis of the results was also carefully performed. After discussing the outcomes of these experiments and analyses, we conclude our paper with suggestions regarding directions of possible future research on this subject.

Recognition of instrument timbres in real polytimbral audio recordings

Automatic recognition of multiple musical instruments in polyphonic and polytimbral music is a difficult task, but often attempted to perform by MIR researchers recently. In papers published so far, the proposed systems were validated mainly on audio data obtained through mixing of isolated sounds of musical instruments. This paper tests recognition of instruments in real recordings, using a recognition system which has multilabel and hierarchical structure. Random forest classifiers were applied to build the system. Evaluation of our model was performed on audio recordings of classical music. The obtained results are shown and discussed in the paper.

Random musical bands playing in random forests

In this paper we investigate the problem of recognizing the full set of instruments playing in a sound mix. Random mixes of 2-5 instruments (out of 14) were created and parameterized to obtain experimental data. Sound samples were taken from 3 audio data sets. For classification purposes, we used a battery of one-instrument sensitive random forest classifiers, and obtained quite good results.

All that jazz in the random forest

In this paper, we address the problem of automatic identification of instruments in audio records, in a frame-by-frame manner. Random forests have been chosen as a classifier. Training data represent sounds of selected instruments which originate from three commonly used repositories, namely McGill University Master Samples, The University of IOWA Musical Instrument Samples, and RWC, as well as from recordings by one of the authors. Testing data represent audio records especially prepared for research purposes, and then carefully labeled (annotated). The experiments on identification of instruments on frame-by-frame basis and the obtained results are presented and discussed in the paper.

Audio-Based Hierarchic Vehicle Classification for Intelligent Transportation Systems

Nowadays almost everybody spends a lot of time commuting and traveling, so we are all very much interested in smooth use of various roads. Also governing bodies are concerned to assure efficient exploitation of the transportation system. The European Union announced a directive on Intelligent Transport Systems in 2010, to ensure that systems integrating information technology with transport engineering are deployed within the Union. In this paper we address automatic classification of vehicle type, based on audio signals only. Hierarchical classification of vehicles is applied, using decision trees, random forests, artificial neural networks, and support vector machines. A dedicated feature set is proposed, based on spectral ranges best separating the target classes. We show that longer analyzing frames yield better results, and a set of binary classifiers performs better than a single multi-class classifier.

Spectral Features for Audio Based Vehicle Identification

In this paper we address automatic vehicle identification based on audio information. Such data are complicated, as they depend on vehicle type, tires, speed and its change. In our previous research we designed a feature set for selected vehicle classes, discriminating pairs of classes. Now, we decided to expand the feature vector and find the best feature set (mainly based on spectral descriptors), possibly representative for each investigated vehicle category, which can be applied to a bigger data set, with more classes. The paper also shows problems related to vehicles classification, which is detailed in official documents by national authority for issues related to the national road system, but simplified for automatic identification purposes. Experiments on audio-based vehicle type identification are presented and conclusions are shown.

Mining audio data for multiple instrument recognition in classical music

This paper addresses the problem of identification of multiple musical instruments in polyphonic recordings of classical music. A set of binary random forests was used as a classifier, and each random forest was trained to recognize the target class of sounds. Training data were prepared in two versions, one based on single sounds and their mixes, and the other containing also sound frames taken from classical music recordings. The experiments on identification of multiple instrument sounds in recordings are presented, and their results are discussed in this paper.

Spectral features for audio based vehicle and engine classification

In this paper we address automatic vehicle and engine identification based on audio information. Such data depend on many factors, including vehicle type, tires, speed and its change, as well as road type. In our previous research we designed a feature set for selected vehicle classes, discriminating pairs of classes. Later, we decided to expand the feature vector and find the best feature set (mainly based on spectral descriptors), possibly representative for each investigated vehicle category, which can be applied to a bigger data set, with more classes. The experiments were performed first on on-road recordings, and then continued with test bench (dyno) recordings. The paper also shows problems related to vehicles classification, which is detailed in official documents by national authority for issues related to the national road system, but simplified for automatic identification purposes. Experiments on audio-based vehicle type and engine type identification are presented and conclusions are shown.