{"type":"report","_id":"11926","oa":"1","author":[{"last_name":"Walter","full_name":"Walter, Oliver","first_name":"Oliver"},{"full_name":"Schmalenstroeer, Joerg","first_name":"Joerg","last_name":"Schmalenstroeer","id":"460"},{"full_name":"Haeb-Umbach, Reinhold","first_name":"Reinhold","id":"242","last_name":"Haeb-Umbach"}],"user_id":"44006","date_created":"2019-07-12T05:30:55Z","citation":{"apa":"Walter, O., Schmalenstroeer, J., &#38; Haeb-Umbach, R. (2013). <i>A Novel Initialization Method for Unsupervised Learning of Acoustic Patterns in Speech (FGNT-2013-01)</i>.","ama":"Walter O, Schmalenstroeer J, Haeb-Umbach R. <i>A Novel Initialization Method for Unsupervised Learning of Acoustic Patterns in Speech (FGNT-2013-01)</i>.; 2013.","mla":"Walter, Oliver, et al. <i>A Novel Initialization Method for Unsupervised Learning of Acoustic Patterns in Speech (FGNT-2013-01)</i>. 2013.","ieee":"O. Walter, J. Schmalenstroeer, and R. Haeb-Umbach, <i>A Novel Initialization Method for Unsupervised Learning of Acoustic Patterns in Speech (FGNT-2013-01)</i>. 2013.","chicago":"Walter, Oliver, Joerg Schmalenstroeer, and Reinhold Haeb-Umbach. <i>A Novel Initialization Method for Unsupervised Learning of Acoustic Patterns in Speech (FGNT-2013-01)</i>, 2013.","bibtex":"@book{Walter_Schmalenstroeer_Haeb-Umbach_2013, title={A Novel Initialization Method for Unsupervised Learning of Acoustic Patterns in Speech (FGNT-2013-01)}, author={Walter, Oliver and Schmalenstroeer, Joerg and Haeb-Umbach, Reinhold}, year={2013} }","short":"O. Walter, J. Schmalenstroeer, R. Haeb-Umbach, A Novel Initialization Method for Unsupervised Learning of Acoustic Patterns in Speech (FGNT-2013-01), 2013."},"status":"public","title":"A Novel Initialization Method for Unsupervised Learning of Acoustic Patterns in Speech (FGNT-2013-01)","department":[{"_id":"54"}],"abstract":[{"lang":"eng","text":"In this paper we present a novel initialization method for unsupervised learning of acoustic patterns in recordings of continuous speech. The pattern discovery task is solved by dynamic time warping whose performance we improve by a smart starting point selection. This enables a more accurate discovery of patterns compared to conventional approaches. After graph-based clustering the patterns are employed for training hidden Markov models for an unsupervised speech acquisition. By iterating between model training and decoding in an EM-like framework the word accuracy is continuously improved. On the TIDIGITS corpus we achieve a word error rate of about 13 percent by the proposed unsupervised pattern discovery approach, which neither assumes knowledge of the acoustic units nor of the labels of the training data."}],"date_updated":"2022-01-06T06:51:12Z","year":"2013","main_file_link":[{"url":"https://groups.uni-paderborn.de/nt/pubs/2013/WaScHa2013.pdf","open_access":"1"}],"language":[{"iso":"eng"}]}