Using Sequential Information in Polyphonic Sound Event Detection

Huang, Guangpu; Heittola, Toni; Virtanen, Tuomas

doi:10.1109/iwaenc.2018.8521367

Cited by 8 publications

(11 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In SED, this results in letting the RNN learn a language model over the sound events, e.g., which sound events are more likely to happen together and/or in sequence, or how likely is a sound event to keep being active, given the previous activity of the sound events. Teacher forcing is different from what was proposed in [13], as the latter approach conditioned the DNN (not the RNN) with the class activities: such an approach yielded poor results, intuitively explained by having y t−1 dominated by the information in X through the sequence of the CNN blocks.…”

Section: Teacher Forcingmentioning

confidence: 93%

“…Finally, we compare our method to the best results presented in [13] which are obtained by employing N-grams as a postprocessing to learn a language model. We report the results of this method on the TUT Sound Events 2016 datasets, as these are the only ones in the corresponding paper that are based on a publicly available dataset.…”

Section: Baselinementioning

confidence: 99%

“…We report the results of this method on the TUT Sound Events 2016 datasets, as these are the only ones in the corresponding paper that are based on a publicly available dataset. It must be noted that in [13] was proposed the usage of y t−1 as extra input features and the usage of CTC, but the results were inferior to the N-grams approach. Specifically, the per frame F1 score was 0.02 and 0.04 lower and ER was 0.02 and 0.15 higher with the usage of y t−1 as an extra input feature and the usage of CTC, respectively, compared to the N-grams approach.…”

Section: Baselinementioning

confidence: 99%

“…Although the results show some improvement with the usage of HMMs, the approach is a hybrid one and it requires a post processing step, which might be limited compared to an non-hybrid, DNN-based approach. In [12] and [13], the connectionist temporal classification (CTC) [14] loss function is used for SED: the output of the DNN is modified in order to be used with the CTC. Although the usage of CTC seems to be promising, CTC needs modification in order to be used for SED, it is a complicated criterion to employ, and it was developed to solve the problem where there is no frame-to-frame alignment between the input and output sequences [14].…”

Section: Introductionmentioning

confidence: 99%

“…Thus, there might be the case that using a different method for SED language modelling could provide better results than CTC. Finally, in [13], the authors also employ N-grams, which require pre and post processing stages, and use the class activities as extra input features. However, the latter approach did not perform better than a baseline which did not employ any language model.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Language Modelling for Sound Event Detection with Teacher Forcing and Scheduled Sampling

Drossos¹,

Gharib²,

Magron³

et al. 2019

Proceedings of the Detection and Classification of Acoustic Scenes And Events 2019 Workshop (DCASE2019)

Self Cite

View full text Add to dashboard Cite

A sound event detection (SED) method typically takes as an input a sequence of audio frames and predicts the activities of sound events in each frame. In real-life recordings, the sound events exhibit some temporal structure: for instance, a "car horn" will likely be followed by a "car passing by". While this temporal structure is widely exploited in sequence prediction tasks (e.g., in machine translation), where language models (LM) are exploited, it is not satisfactorily modeled in SED. In this work we propose a method which allows a recurrent neural network (RNN) to learn an LM for the SED task. The method conditions the input of the RNN with the activities of classes at the previous time step. We evaluate our method using F1 score and error rate (ER) over three different and publicly available datasets; the TUT-SED Synthetic 2016 and the TUT Sound Events 2016 and 2017 datasets. The obtained results show an increase of 6% and 3% at the F1 (higher is better) and a decrease of 3% and 2% at ER (lower is better) for the TUT Sound Events 2016 and 2017 datasets, respectively, when using our method. On the contrary, with our method there is a decrease of 10% at F1 score and an increase of 11% at ER for the TUT-SED Synthetic 2016 dataset.

show abstract

Section: Teacher Forcingmentioning

confidence: 93%

Section: Baselinementioning

confidence: 99%

Section: Baselinementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations