CONF valente:Interspeech:2007/IDIAP Hierarchical Neural Networks Feature Extraction for LVCSR system Valente, Fabio Vepa, Jithendra Plahl, Christian Gollan, Christian Hermansky, Hynek Schlüter, Ralf EXTERNAL https://publications.idiap.ch/attachments/papers/2007/valente-Interspeech-2007.pdf PUBLIC https://publications.idiap.ch/index.php/publications/showcite/valente:rr07-08 Related documents Interspeech 2007 2007 IDIAP-RR 07-08 This paper investigates the use of a hierarchy of Neural Networks for performing data driven feature extraction. Two different hierarchical structures based on long and short temporal context are considered. Features are tested on two different LVCSR systems for Meetings data (RT05 evaluation data) and for Arabic Broadcast News (BNAT05 evaluation data). The hierarchical NNs outperforms the single NN features consistently on different type of data and tasks and provides significant improvements w.r.t. respective baselines systems. Best result is obtained when different time resolutions are used at different level of the hierarchy. REPORT valente:rr07-08/IDIAP Hierarchical Neural Networks Feature Extraction for LVCSR system Valente, Fabio Vepa, Jithendra Plahl, Christian Gollan, Christian Hermansky, Hynek Schlüter, Ralf EXTERNAL https://publications.idiap.ch/attachments/reports/2007/valente-idiap-rr-07-08.pdf PUBLIC Idiap-RR-08-2007 2007 IDIAP Submitted for publication This paper investigates the use of a hierarchy of Neural Networks for performing data driven feature extraction. Two different hierarchical structures based on long and short temporal context are considered. Features are tested on two different LVCSR systems for Meetings data (RT05 evaluation data) and for Arabic Broadcast News (BNAT05 evaluation data). The hierarchical NNs outperforms the single NN features consistently on different type of data and tasks and provides significant improvements w.r.t. respective baselines systems. Best result is obtained when different time resolutions are used at different level of the hierarchy.