{"id":{"repo_id":"oldenburg","oai_identifier":"oai:oops.uni-oldenburg.de:282"},"canonical_url":"https://search.dev.ndltd.org/etd/oldenburg/oai:oops.uni-oldenburg.de:282","repository":{"repo_id":"oldenburg","name":"Carl von Ossietzky Universität Oldenburg","base_url":"http://oops.uni-oldenburg.de/cgi/oai2"},"display":{"title":"Robust speech recognition based on spectro-temporal processing","abstract":"In this thesis, novelle spectro-temporal feature extraction techniques are evaluated for enhancing the robustness of automatic speech recognition systems (ASR) in adverse acoustical conditions. Recent physiological and psychoacoustical findings indicate that spectro-temporal processing plays an important role in human speech perception. Therefore, sigma-pi cells and Gabor filter functions are investigated as secondary feature extraction methods based on spectro-temporal representation. Especially the Gabor features are versatile enough to include cepstral features and purely temporal filtering as special cases, while additionally aiming at combined spectro-temporal modulations. A data driven feature selection method is applied for feature set optimization. For small vocabularies, both types of features are shown to increase the robustness of ASR systems. Sigma-pi cells also allow for estimating the speech-to-noise ratio of an input signal solely based on low spectro-temporal modulation. The Gabor based Tandem feature sets increase the performance of the Qualcomm-ICSI-OGI system for the Aurora task, when concatenating the two streams.","abstract_html":"In this thesis, novelle spectro-temporal feature extraction techniques are evaluated for enhancing the robustness of automatic speech recognition systems (ASR) in adverse acoustical conditions. Recent physiological and psychoacoustical findings indicate that spectro-temporal processing plays an important role in human speech perception. Therefore, sigma-pi cells and Gabor filter functions are investigated as secondary feature extraction methods based on spectro-temporal representation. Especially the Gabor features are versatile enough to include cepstral features and purely temporal filtering as special cases, while additionally aiming at combined spectro-temporal modulations. A data driven feature selection method is applied for feature set optimization. For small vocabularies, both types of features are shown to increase the robustness of ASR systems. Sigma-pi cells also allow for estimating the speech-to-noise ratio of an input signal solely based on low spectro-temporal modulation. The Gabor based Tandem feature sets increase the performance of the Qualcomm-ICSI-OGI system for the Aurora task, when concatenating the two streams.","abstract_has_math":false,"creators":["Kleinschmidt, Michael"],"institution":"Universität Oldenburg","degree_name":null,"degree_level":"thesis.doctoral","degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2002,"date_issued":"2002-09-05","date_published":"2002-09-05","updated_at":"2026-07-27T20:27:18Z","subjects":["[Keine Schlagwörter von Autor/in vergeben.]"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://oops.uni-oldenburg.de/282","outbound_label":"Repository record","outbound_source":"source_url"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:creator","label":"Author","values":["Kleinschmidt, Michael"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:publisher","label":"Institution","values":["BIS der Universität Oldenburg"]},{"key":"dc:type","label":"Dc Type","values":["doctoralThesis"]},{"key":"thesis:degree_level","label":"Degree Level","values":["thesis.doctoral"]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["Universität Oldenburg"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["[Keine Schlagwörter von Autor/in vergeben.]"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["In this thesis, novelle spectro-temporal feature extraction techniques are evaluated for enhancing the robustness of automatic speech recognition systems (ASR) in adverse acoustical conditions. Recent physiological and psychoacoustical findings indicate that spectro-temporal processing plays an important role in human speech perception. Therefore, sigma-pi cells and Gabor filter functions are investigated as secondary feature extraction methods based on spectro-temporal representation. Especially the Gabor features are versatile enough to include cepstral features and purely temporal filtering as special cases, while additionally aiming at combined spectro-temporal modulations. A data driven feature selection method is applied for feature set optimization. For small vocabularies, both types of features are shown to increase the robustness of ASR systems. Sigma-pi cells also allow for estimating the speech-to-noise ratio of an input signal solely based on low spectro-temporal modulation. The Gabor based Tandem feature sets increase the performance of the Qualcomm-ICSI-OGI system for the Aurora task, when concatenating the two streams.","In dieser Dissertation werden neuartige spektro-temporale Merkmale untersucht, die einer Verbesserung der Robustheit automatischer Spracherkennungssysteme unter ungünstigen akustischen Bedingungen dienen sollen. Ergebnisse physiologischer und psychoakustischer Arbeiten weisen auf eine wichtige Rolle spektro-temporaler Verarbeitung bei der Sprachwahrnehmung des Menschen hin. Daher werden Sigma-pi Zellen und Gabor Filter als Methoden zur Extraktion sekundärer Merkmale auf Basis einer spektro-temporalen Repräsentation evaluiert. Insbesondere die Gabor Merkmale beinhalten das Cepstrum sowie eine rein zeitliche Filterung als Spezialfälle, wobei darüber hinaus auf spektro-temporale Modulationen gezielt wird. Eine datenbasierte Methode zur Merkmalsselektion wird zur Optimierung der Merkmalssätze verwendet. Beide Typen von Merkmalen zeigen eine erhöhte Robustheit bei Experimenten mit kleinen Wortschätzen. Sigma-pi Zellen erlauben zudem eine Schätzung des Sprach-zu-Rausch Abstandes des Eingangssignals allein aufgrund kleiner spektro-temporaler Modulationen. Durch Anhängen von Gabor-basierten Merkmalen kann die Erkennungsleistung des Qualcomm-ICSI-OGI Erkenners im Aurora Experiment weiter verbessert werden."]},{"key":"dc:format.medium","label":"Dc Format Medium","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Robust speech recognition based on spectro-temporal processing"]}]}],"canonical_facts":{"dc:creator":["Kleinschmidt, Michael"],"dc:description.abstract":["In this thesis, novelle spectro-temporal feature extraction techniques are evaluated for enhancing the robustness of automatic speech recognition systems (ASR) in adverse acoustical conditions. Recent physiological and psychoacoustical findings indicate that spectro-temporal processing plays an important role in human speech perception. Therefore, sigma-pi cells and Gabor filter functions are investigated as secondary feature extraction methods based on spectro-temporal representation. Especially the Gabor features are versatile enough to include cepstral features and purely temporal filtering as special cases, while additionally aiming at combined spectro-temporal modulations. A data driven feature selection method is applied for feature set optimization. For small vocabularies, both types of features are shown to increase the robustness of ASR systems. Sigma-pi cells also allow for estimating the speech-to-noise ratio of an input signal solely based on low spectro-temporal modulation. The Gabor based Tandem feature sets increase the performance of the Qualcomm-ICSI-OGI system for the Aurora task, when concatenating the two streams.","In dieser Dissertation werden neuartige spektro-temporale Merkmale untersucht, die einer Verbesserung der Robustheit automatischer Spracherkennungssysteme unter ungünstigen akustischen Bedingungen dienen sollen. Ergebnisse physiologischer und psychoakustischer Arbeiten weisen auf eine wichtige Rolle spektro-temporaler Verarbeitung bei der Sprachwahrnehmung des Menschen hin. Daher werden Sigma-pi Zellen und Gabor Filter als Methoden zur Extraktion sekundärer Merkmale auf Basis einer spektro-temporalen Repräsentation evaluiert. Insbesondere die Gabor Merkmale beinhalten das Cepstrum sowie eine rein zeitliche Filterung als Spezialfälle, wobei darüber hinaus auf spektro-temporale Modulationen gezielt wird. Eine datenbasierte Methode zur Merkmalsselektion wird zur Optimierung der Merkmalssätze verwendet. Beide Typen von Merkmalen zeigen eine erhöhte Robustheit bei Experimenten mit kleinen Wortschätzen. Sigma-pi Zellen erlauben zudem eine Schätzung des Sprach-zu-Rausch Abstandes des Eingangssignals allein aufgrund kleiner spektro-temporaler Modulationen. Durch Anhängen von Gabor-basierten Merkmalen kann die Erkennungsleistung des Qualcomm-ICSI-OGI Erkenners im Aurora Experiment weiter verbessert werden."],"dc:format.medium":["application/pdf"],"dc:publisher":["BIS der Universität Oldenburg"],"dc:subject":["[Keine Schlagwörter von Autor/in vergeben.]"],"dc:title":["Robust speech recognition based on spectro-temporal processing"],"dc:type":["doctoralThesis"],"thesis:degree_level":["thesis.doctoral"],"thesis:institution_name":["Universität Oldenburg"]},"updated_at":"2026-07-27T20:27:18Z"}