Abstract
Sound Event Detection is a task with a rising relevance over the recent years in the field of audio signal processing, due to the creation of specific datasets such as Google AudioSet or DESED (Domestic Environment Sound Event Detection) and the introduction of competitive evaluations like the DCASE Challenge (Detection and Classification of Acoustic Scenes and Events). The different categories of acoustic events can present diverse temporal and spectral characteristics. However, most approaches use a fixed time-frequency resolution to represent the audio segments. This work proposes a multi-resolution analysis for feature extraction in Sound Event Detection, hypothesizing that different resolutions can be more adequate for the detection of different sound event categories, and that combining the information provided by multiple resolutions could improve the performance of Sound Event Detection systems. Experiments are carried out over the DESED dataset in the context of the DCASE 2020 Challenge, concluding that the combination of up to 5 resolutions allows a neural network-based system to obtain better results than single-resolution models in terms of event-based F1-score in every event category and in terms of PSDS (Polyphonic Sound Detection Score). Furthermore, we analyze the impact of score thresholding in the computation of F1-score results, finding that the standard value of 0.5 is suboptimal and proposing an alternative strategy based in the use of a specific threshold for each event category, which obtains further improvements in performance.
Highlights
U NDERSTANDING the acoustic environment is an ongoing challenge for artificial intelligence which has motivated several research fields
Taking into account that the BS resolution point coincides with the baseline system of DCASE Challenge 2020 Task 4, the results obtained using this resolution constitute the common benchmark for the aforementioned task
The results of the single-resolution models are presented in Table 3 as the mean and the standard deviation of the F1-scores obtained with the five trainings
Summary
U NDERSTANDING the acoustic environment is an ongoing challenge for artificial intelligence which has motivated several research fields. The use of two different resolutions has been proposed to improve automatic speech recognition in reverberant scenarios [23], in which a wide-context window gives information about the acoustic environment and reverberation, whereas a narrow-context window provides finer detail about the content of the speech signal This is possible due to the existence of a tradeoff between time resolution and frequency resolution in the extraction of Fast Fourier Transform-based audio features [24] such as the mel-spectrogram, which is the base for the analysis proposed in this work. The proposed analysis is tested using a state-of-the-art system, the baseline for DCASE 2020 Challenge Task 4 “Detection and Separation of Sound Events in Domestic Environments” [25] The aim of this challenge is to make use of unlabeled and weakly-labeled recordings, together with strongly-labeled synthetic audio clips, to train systems that predict the temporal locations of ten different event categories in audio recordings.
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
