Synthetic data generation is an interesting ara of research but i have difficulties finding articles and textbooks about the topic. I want an idea about definitions and framework for automatic synthetic data generations in any area, particullary on sound analysis.
I am not sure if the following is useful to you, but in 2008 we wrote a paper on simulating labelled (medical) image data motivated by the need for large datasets for training machine learning and for validating algorithm: http://www.cs.sfu.ca/~hamarneh/ecopy/miccai2008b.pdf You can also see http://vascusynth.cs.sfu.ca/ where we simulated branching tubular structures (such as vascular trees).
Thank you for your answer, Dr. Hamerneh. I see that the synthetic data generation is focused on medical applications like your article, proof of methods in computer systems confiability and image analysis. Your approach is interesting, thank you for your help.
As you focus on sound analysis, you may find interesting nowadays state-of-the-art technique for improving speech recognition acoustic models by augmenting training data using with speed, frequency and tempo warping, which is indirectly related to synthetic data generation ( http://www.danielpovey.com/files/2015_interspeech_augmentation.pdf )
Another example of creating synthetic data is block mixing that seems to be useful in polyphonic acoustic event detection: https://arxiv.org/pdf/1604.00861.pdf
We also tried the latter in recent DCASE challenge gaining a little bit on F-measure, but not as large as the author of previous work: https://www.researchgate.net/publication/306118891_DCASE_2016_Sound_Event_Detection_System_Based_on_Convolutional_Neural_Network
Technical Report DCASE 2016 Sound Event Detection System Based on Convolution...
Thank you for your help Dr. Gorin. The strategies presented in the article of No are the ideas that I was looking for.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence