{"keyword":["voice activity detection","speech activity detection","neural network","statistical speech processing"],"file":[{"file_name":"ms.pdf","creator":"jensheit","file_size":3871374,"date_updated":"2020-12-11T12:36:37Z","file_id":"20699","access_level":"closed","success":1,"date_created":"2020-12-11T12:36:37Z","content_type":"application/pdf","relation":"main_file"}],"ddc":["000"],"abstract":[{"text":"In recent years time domain speech separation has excelled over frequency domain separation in single channel scenarios and noise-free environments. In this paper we dissect the gains of the time-domain audio separation network (TasNet) approach by gradually replacing components of an utterance-level permutation invariant training (u-PIT) based separation system in the frequency domain until the TasNet system is reached, thus blending components of frequency domain approaches with those of time domain approaches. Some of the intermediate variants achieve comparable signal-to-distortion ratio (SDR) gains to TasNet, but retain the advantage of frequency domain processing: compatibility with classic signal processing tools such as frequency-domain beamforming and the human interpretability of the masks. Furthermore, we show that the scale invariant signal-to-distortion ratio (si-SDR) criterion used as loss function in TasNet is related to a logarithmic mean square error criterion and that it is this criterion which contributes most reliable to the performance advantage of TasNet. Finally, we critically assess which gains in a noise-free single channel environment generalize to more realistic reverberant conditions.","lang":"eng"}],"author":[{"first_name":"Jens","id":"27643","full_name":"Heitkaemper, Jens","last_name":"Heitkaemper"},{"last_name":"Jakobeit","full_name":"Jakobeit, Darius","first_name":"Darius"},{"full_name":"Boeddeker, Christoph","last_name":"Boeddeker","first_name":"Christoph","id":"40767"},{"full_name":"Drude, Lukas","last_name":"Drude","first_name":"Lukas"},{"first_name":"Reinhold","id":"242","full_name":"Haeb-Umbach, Reinhold","last_name":"Haeb-Umbach"}],"status":"public","year":"2020","_id":"20504","file_date_updated":"2020-12-11T12:36:37Z","type":"conference","date_created":"2020-11-25T14:56:53Z","publication":"ICASSP 2020 Virtual Barcelona Spain","project":[{"name":"PC2: Computing Resources Provided by the Paderborn Center for Parallel Computing","_id":"52"}],"department":[{"_id":"54"}],"user_id":"40767","citation":{"apa":"Heitkaemper, J., Jakobeit, D., Boeddeker, C., Drude, L., &#38; Haeb-Umbach, R. (2020). Demystifying TasNet: A Dissecting Approach. <i>ICASSP 2020 Virtual Barcelona Spain</i>.","ieee":"J. Heitkaemper, D. Jakobeit, C. Boeddeker, L. Drude, and R. Haeb-Umbach, “Demystifying TasNet: A Dissecting Approach,” 2020.","mla":"Heitkaemper, Jens, et al. “Demystifying TasNet: A Dissecting Approach.” <i>ICASSP 2020 Virtual Barcelona Spain</i>, 2020.","ama":"Heitkaemper J, Jakobeit D, Boeddeker C, Drude L, Haeb-Umbach R. Demystifying TasNet: A Dissecting Approach. In: <i>ICASSP 2020 Virtual Barcelona Spain</i>. ; 2020.","chicago":"Heitkaemper, Jens, Darius Jakobeit, Christoph Boeddeker, Lukas Drude, and Reinhold Haeb-Umbach. “Demystifying TasNet: A Dissecting Approach.” In <i>ICASSP 2020 Virtual Barcelona Spain</i>, 2020.","bibtex":"@inproceedings{Heitkaemper_Jakobeit_Boeddeker_Drude_Haeb-Umbach_2020, title={Demystifying TasNet: A Dissecting Approach}, booktitle={ICASSP 2020 Virtual Barcelona Spain}, author={Heitkaemper, Jens and Jakobeit, Darius and Boeddeker, Christoph and Drude, Lukas and Haeb-Umbach, Reinhold}, year={2020} }","short":"J. Heitkaemper, D. Jakobeit, C. Boeddeker, L. Drude, R. Haeb-Umbach, in: ICASSP 2020 Virtual Barcelona Spain, 2020."},"date_updated":"2022-01-13T08:47:32Z","has_accepted_license":"1","quality_controlled":"1","language":[{"iso":"eng"}],"title":"Demystifying TasNet: A Dissecting Approach"}