Convolving Gaussian Kernels for RNN-Based Beat Tracking. Cheng, T., Fukayama, S., & Goto, M. In 2018 26th European Signal Processing Conference (EUSIPCO), pages 1905-1909, Sep., 2018. Paper doi abstract bibtex Because of an ability of modelling context information, Recurrent Neural Networks (RNNs) or bi-directional RNNs (BRNNs) have been used for beat tracking with good performance. However, there are two problems associated with RNN-based beat tracking. The first problem is the imbalanced data: usually only around 2% frames are labelled as `beat'. The second one is the disagreement on the precise positions of beats in human annotations or the delay of annotations caused by human tapping. In order to tackle these problems, we propose to convolve the original ground truth with a Gaussian kernel as the target output of the network for a more robust training. We conduct a comparison experiment using five different Gaussian kernels on five individual datasets. The results on the validation sets show that we can train a better or at least competitive model in a shorter time by using the convolved ground truth with a proper Gaussian kernel.
@InProceedings{8553310,
author = {T. Cheng and S. Fukayama and M. Goto},
booktitle = {2018 26th European Signal Processing Conference (EUSIPCO)},
title = {Convolving Gaussian Kernels for RNN-Based Beat Tracking},
year = {2018},
pages = {1905-1909},
abstract = {Because of an ability of modelling context information, Recurrent Neural Networks (RNNs) or bi-directional RNNs (BRNNs) have been used for beat tracking with good performance. However, there are two problems associated with RNN-based beat tracking. The first problem is the imbalanced data: usually only around 2% frames are labelled as `beat'. The second one is the disagreement on the precise positions of beats in human annotations or the delay of annotations caused by human tapping. In order to tackle these problems, we propose to convolve the original ground truth with a Gaussian kernel as the target output of the network for a more robust training. We conduct a comparison experiment using five different Gaussian kernels on five individual datasets. The results on the validation sets show that we can train a better or at least competitive model in a shorter time by using the convolved ground truth with a proper Gaussian kernel.},
keywords = {audio signal processing;Gaussian processes;recurrent neural nets;bi-directional RNN;recurrent neural networks;Gaussian kernel;convolved ground truth;human annotations;RNN-based beat tracking;Training;Kernel;Standards;Recurrent neural networks;Bidirectional control;Europe;Signal processing},
doi = {10.23919/EUSIPCO.2018.8553310},
issn = {2076-1465},
month = {Sep.},
url = {https://www.eurasip.org/proceedings/eusipco/eusipco2018/papers/1570437485.pdf},
}
Downloads: 0
{"_id":"5Pmv4M6cYCXkaM6YB","bibbaseid":"cheng-fukayama-goto-convolvinggaussiankernelsforrnnbasedbeattracking-2018","authorIDs":[],"author_short":["Cheng, T.","Fukayama, S.","Goto, M."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"firstnames":["T."],"propositions":[],"lastnames":["Cheng"],"suffixes":[]},{"firstnames":["S."],"propositions":[],"lastnames":["Fukayama"],"suffixes":[]},{"firstnames":["M."],"propositions":[],"lastnames":["Goto"],"suffixes":[]}],"booktitle":"2018 26th European Signal Processing Conference (EUSIPCO)","title":"Convolving Gaussian Kernels for RNN-Based Beat Tracking","year":"2018","pages":"1905-1909","abstract":"Because of an ability of modelling context information, Recurrent Neural Networks (RNNs) or bi-directional RNNs (BRNNs) have been used for beat tracking with good performance. However, there are two problems associated with RNN-based beat tracking. The first problem is the imbalanced data: usually only around 2% frames are labelled as `beat'. The second one is the disagreement on the precise positions of beats in human annotations or the delay of annotations caused by human tapping. In order to tackle these problems, we propose to convolve the original ground truth with a Gaussian kernel as the target output of the network for a more robust training. We conduct a comparison experiment using five different Gaussian kernels on five individual datasets. The results on the validation sets show that we can train a better or at least competitive model in a shorter time by using the convolved ground truth with a proper Gaussian kernel.","keywords":"audio signal processing;Gaussian processes;recurrent neural nets;bi-directional RNN;recurrent neural networks;Gaussian kernel;convolved ground truth;human annotations;RNN-based beat tracking;Training;Kernel;Standards;Recurrent neural networks;Bidirectional control;Europe;Signal processing","doi":"10.23919/EUSIPCO.2018.8553310","issn":"2076-1465","month":"Sep.","url":"https://www.eurasip.org/proceedings/eusipco/eusipco2018/papers/1570437485.pdf","bibtex":"@InProceedings{8553310,\n author = {T. Cheng and S. Fukayama and M. Goto},\n booktitle = {2018 26th European Signal Processing Conference (EUSIPCO)},\n title = {Convolving Gaussian Kernels for RNN-Based Beat Tracking},\n year = {2018},\n pages = {1905-1909},\n abstract = {Because of an ability of modelling context information, Recurrent Neural Networks (RNNs) or bi-directional RNNs (BRNNs) have been used for beat tracking with good performance. However, there are two problems associated with RNN-based beat tracking. The first problem is the imbalanced data: usually only around 2% frames are labelled as `beat'. The second one is the disagreement on the precise positions of beats in human annotations or the delay of annotations caused by human tapping. In order to tackle these problems, we propose to convolve the original ground truth with a Gaussian kernel as the target output of the network for a more robust training. We conduct a comparison experiment using five different Gaussian kernels on five individual datasets. The results on the validation sets show that we can train a better or at least competitive model in a shorter time by using the convolved ground truth with a proper Gaussian kernel.},\n keywords = {audio signal processing;Gaussian processes;recurrent neural nets;bi-directional RNN;recurrent neural networks;Gaussian kernel;convolved ground truth;human annotations;RNN-based beat tracking;Training;Kernel;Standards;Recurrent neural networks;Bidirectional control;Europe;Signal processing},\n doi = {10.23919/EUSIPCO.2018.8553310},\n issn = {2076-1465},\n month = {Sep.},\n url = {https://www.eurasip.org/proceedings/eusipco/eusipco2018/papers/1570437485.pdf},\n}\n\n","author_short":["Cheng, T.","Fukayama, S.","Goto, M."],"key":"8553310","id":"8553310","bibbaseid":"cheng-fukayama-goto-convolvinggaussiankernelsforrnnbasedbeattracking-2018","role":"author","urls":{"Paper":"https://www.eurasip.org/proceedings/eusipco/eusipco2018/papers/1570437485.pdf"},"keyword":["audio signal processing;Gaussian processes;recurrent neural nets;bi-directional RNN;recurrent neural networks;Gaussian kernel;convolved ground truth;human annotations;RNN-based beat tracking;Training;Kernel;Standards;Recurrent neural networks;Bidirectional control;Europe;Signal processing"],"metadata":{"authorlinks":{}},"downloads":0},"bibtype":"inproceedings","biburl":"https://raw.githubusercontent.com/Roznn/EUSIPCO/main/eusipco2018url.bib","creationDate":"2021-02-13T15:38:40.384Z","downloads":0,"keywords":["audio signal processing;gaussian processes;recurrent neural nets;bi-directional rnn;recurrent neural networks;gaussian kernel;convolved ground truth;human annotations;rnn-based beat tracking;training;kernel;standards;recurrent neural networks;bidirectional control;europe;signal processing"],"search_terms":["convolving","gaussian","kernels","rnn","based","beat","tracking","cheng","fukayama","goto"],"title":"Convolving Gaussian Kernels for RNN-Based Beat Tracking","year":2018,"dataSources":["yiZioZximP7hphDpY","iuBeKSmaES2fHcEE9"]}