|
|
Beat tracking systems for music recordings
Staňková, Karolína ; Miklánek, Štěpán (referee) ; Ištvánek, Matěj (advisor)
This master thesis deals with systems for detecting rhythmic structures of music recordings. The field of Music Information Retrieval (MIR) allows us to examine the harmonic and tonal properties of music, rhythm, tempo, etc., and has uses in academic and commercial sphere. Various algorithms are used in the detection of rhythmic structures. However, today, most new methods use neural networks. This work aims to summarize the current research results of systems for detecting beats and tempo, to describe methods of calculating and evaluating the parameters of music recordings, and to implement a program that allows comparison of available detection systems. The result of the work is a script in the Python language, which uses six different systems to detect the rhythmic structure of test recordings. It then checks the outputs of the algorithms according to the given reference and compares the given systems with each other using several evaluation values. It uses two datasets as a reference—one of them is publicly available and the other was created by the author of this thesis (including annotations, i.e., reference beat times, for the sample recordings). The program allows user to see the results in graphs and play any of the sample recordings with detected beat times.
|
|
|
Tempo detector based on a neural network
Suchánek, Tomáš ; Smékal, Zdeněk (referee) ; Ištvánek, Matěj (advisor)
This Master’s thesis deals with beat tracking systems, whose functionality is based on neural networks. It describes the structure of these systems and how the signal is processed in their individual blocks. Emphasis is then placed on recurrent and temporal convolutional networks, which by they nature can effectively detect tempo and beats in audio recordings. The selected methods, network architectures and their modifications are then implemented within a comprehensive detection system, which is further tested and evaluated through a cross-validation process on a genre-diverse data-set. The results show that the system, with proposed temporal convolutional network architecture, produces comparable results with foreign publications. For example, within the SMC dataset, it proved to be the most successful, on the contrary, in the case of other datasets it was slightly below the accuracy of state-of-the-art systems. In addition,the proposed network retains low computational complexity despite increased number of internal parameters.
|
|
|
Beat Tracking: Is 441 kHz Really Needed?
Ištvánek, Matěj ; Miklánek, Štěpán
Beat tracking is essential in music informationretrieval, with applications ranging from music analysis and automaticplaylist generation to beat-synchronized effects. In recentyears, deep learning methods, usually inspired by well-knownarchitectures, outperformed other beat tracking algorithms. Thecurrent state-of-the-art offline beat tracking systems utilize temporalconvolutional and recurrent networks. Most systems use aninput sampling rate of 44.1 kHz. In this paper, we retrain multipleversions of state-of-the-art temporal convolutional networks withdifferent input sampling rates while keeping the time resolutionby changing the frame size parameter. Furthermore, we evaluateall models using standard metrics. As the main contribution,we show that decreasing the input audio recording samplingfrequency up to 5 kHz preserves most of the accuracy, and insome cases, even slightly outperforms the standard approach.
|
|
|
The Application Of Tempo Calculation For Musicological Purposes
Istvanek, Matej
Beat tracking systems capture time positions of beats within digital recordings. Theyare used, for example, in streaming portals, but applications in the musicological analysis are oftenneglected. In this article, two different methods of beat tracking systems are tested—conventionaland the state-of-the-art—on the specific motif of a string quartet music, which is one of the mostcomplex tasks for beat detectors in general. The aim here is to determine which system is better formusicology purposes. This often involves determining not only the position of individual beats andestimating the tempo but also the accuracy of determining their number. Evaluation analysis maybe suitable for comparing the accuracy of detectors, but may not necessarily reflect the requirementsof musicological analysis. The results of selected detectors show that a system based on a recurrentneural network seems to be the most suitable.
|
|
|
Beat Tracking System Based On A Neural Network
Suchánek, Tomáš
This thesis deals with systems for tempo and beat detection in music recordings, whosefunctionality is based on neural networks. The basic structure of such systems is briefly described andthe emphasis is then placed on a comparison of recurrent and temporal convolutional networks, whichhave proven to be the most suitable for this task. The main outcome of this work is then proposaland comparison of modified temporal convolutional network with other state-of-the-art networks ina beat tracking system. The results suggest that simplification in existing architectures could benefitfrom faster training times, while it maintains or slightly improves the accuracy of a detection system.
|
|
|
Comparison And Evaluation System For Beat Tracking Algorithms
Staňková, Karolína
This work deals with systems for detecting rhythmic structures of music recordings. Thefield of retrieving information from music (MIR) is developing rapidly and has more accurate resultsthan ever before. It allows us to examine the harmonic and tonal properties of music, rhythm, tempo,etc. Various algorithms are used in the detection of rhythmic structures. However, today, most newmethods use neural networks. The thesis aims to summarize the current research results of systemsfor detecting music times and tempo in MIR, to describe methods of calculating and evaluating theparameters of music recordings, and to implement a program that allows comparison of availabledetection systems. The result of the work is a script in the Python language, which uses five differentsystems to detect the rhythmic structure of test recordings. It then checks the outputs of the algorithmsaccording to the given reference and compares the given systems with each other using severalevaluation quantities.
|
|
|
Tempo detector based on a neural network
Suchánek, Tomáš ; Smékal, Zdeněk (referee) ; Ištvánek, Matěj (advisor)
This Master’s thesis deals with beat tracking systems, whose functionality is based on neural networks. It describes the structure of these systems and how the signal is processed in their individual blocks. Emphasis is then placed on recurrent and temporal convolutional networks, which by they nature can effectively detect tempo and beats in audio recordings. The selected methods, network architectures and their modifications are then implemented within a comprehensive detection system, which is further tested and evaluated through a cross-validation process on a genre-diverse data-set. The results show that the system, with proposed temporal convolutional network architecture, produces comparable results with foreign publications. For example, within the SMC dataset, it proved to be the most successful, on the contrary, in the case of other datasets it was slightly below the accuracy of state-of-the-art systems. In addition,the proposed network retains low computational complexity despite increased number of internal parameters.
|
|
|
Beat tracking systems for music recordings
Staňková, Karolína ; Miklánek, Štěpán (referee) ; Ištvánek, Matěj (advisor)
This master thesis deals with systems for detecting rhythmic structures of music recordings. The field of Music Information Retrieval (MIR) allows us to examine the harmonic and tonal properties of music, rhythm, tempo, etc., and has uses in academic and commercial sphere. Various algorithms are used in the detection of rhythmic structures. However, today, most new methods use neural networks. This work aims to summarize the current research results of systems for detecting beats and tempo, to describe methods of calculating and evaluating the parameters of music recordings, and to implement a program that allows comparison of available detection systems. The result of the work is a script in the Python language, which uses six different systems to detect the rhythmic structure of test recordings. It then checks the outputs of the algorithms according to the given reference and compares the given systems with each other using several evaluation values. It uses two datasets as a reference—one of them is publicly available and the other was created by the author of this thesis (including annotations, i.e., reference beat times, for the sample recordings). The program allows user to see the results in graphs and play any of the sample recordings with detected beat times.
|
|
|
Music Visualization in 3D
Vincena, Petr ; Rittig, Tobias (advisor) ; Hajič, Jan (referee)
Music is an important part of our lives and music emotion recognition is an important field of study with many applications. In this work, we focus on music features extraction and subsequent creation of music emotion recognition system. We extend the work of Hun at al. [2009] in order to create a feed-forward neural network model that predicts emotions from music which we can use in visualization. We confirm the results and discuss their impact. We also provide a C# project to extract several features from music that can be used in Unity together with the demonstration scene that visualize these features. 1
|
|
|
Enhancement Of Global Tempo Computation In Beat Tracking System Based On Teager-Kaiser Energy Operator
Ištvánek, Matěj
Beat detection systems and onset detections are used in music information retrieval (MIR) research field for the calculation of the global tempo (GT) and beat positions in audio recordings. The aim of this article is to introduce the enhancement of the onset detector and therefore the beat tracking system. The enhancement is based on the Teager-Kaiser energy operator (TKEO), which is used in pre-processing stage before the onset computation. The proposed method is firstly evaluated in terms of ability to estimate GT of a given audio track and then it is tested on the string quartet database. Results suggest that the TKEO could improve accuracy of GT estimation. Proposed beat tracking system could be used for analysis of interpretation changes in string quartet music.
|
guest ::
