CONVOLUTIONAL NEURAL NETWORK WITH MULTIPLE-WIDTH FREQUENCY-DELTA DATA AUGMENTATION FOR ACOUSTIC SCENE CLASSIFICATION

In this paper, we apply convolutional neural network on acoustic scene classification task of DCASE 2016. We propose multiwidth frequency-delta data augmentation which uses static melspectrogram as well as frequency-delta features as individual examples with same labels for the network input, and the experimental result shows that this method significantly improves the performance compare to the case of using static mel-spectrogram input only. In addition, we propose folded mean aggregation, which first multiplies output probabilities of static and delta augmentation data from the same window first prior to audio clip-wise aggregation, and we found that this method reduces the error rate further. The system exhibited a classification accuracy of 0.831 on the development set and 0.846 on the evaluation set.

[1]  Andrew L. Maas.  Rectifier Nonlinearities Improve Neural Network Acoustic Models , 2013 .

[2]  Dan Stowell,et al.  Detection and classification of acoustic scenes and events: An IEEE AASP challenge , 2013, 2013 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics.

[3]  Sebastian Böck,et al.  Improved musical onset detection with Convolutional Neural Networks , 2014, 2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).

[4]  Mark D. Plumbley,et al.  Acoustic Scene Classification: Classifying environments from the sounds they produce , 2014, IEEE Signal Processing Magazine.

[5]  Andrew Zisserman,et al.  Very Deep Convolutional Networks for Large-Scale Image Recognition , 2014, ICLR.

[6]  Tuomas Virtanen,et al.  TUT database for acoustic scene classification and sound event detection , 2016, 2016 24th European Signal Processing Conference (EUSIPCO).

[7]  Guigang Zhang,et al.  Deep Learning , 2016, Int. J. Semantic Comput..

[8]  Jae-Hun Kim,et al.  Deep Convolutional Neural Networks for Predominant Instrument Recognition in Polyphonic Music , 2016, IEEE/ACM Transactions on Audio, Speech, and Language Processing.