|
|
Query-by-Example Spoken Term Detection
Fapšo, Michal ; Matoušek, Jindřich (oponent) ; Metze, Florian (oponent) ; Černocký, Jan (vedoucí práce)
This thesis investigates query-by-example (QbE) spoken term detection (STD). Queries are entered in their spoken form and searched for in a pool of recorded spoken utterances, providing a list of detections with their scores and timing. We describe, analyze and compare three different approaches to QbE STD, in various language-dependent and language-independent setups with diverse audio conditions, searching for a single example and five examples per query. For our experiments we used Czech, Hungarian, English and Levantine data and for each of the languages we trained a 3-state phone posterior estimator. This gave us 16 possible combinations of the evaluation language and the language of the posterior estimator, out of which 4 combinations were language-dependent and 12 were language-independent. All QbE systems were evaluated on the same data and the same features, using the metrics: non-pooled Figure-of-Merit and our proposed utterrance-normalized non-pooled Figure-of-Merit, which provided us with relevant data for the comparison of these QbE approaches and for gaining a better insight into their behavior. QbE approaches presented in this work are: sequential statistical modeling (GMM/HMM), template matching of features (DTW) and matching of phone lattices (WFST). To compare the performance of QbE approaches with the common query-by-text STD systems, for language-dependent setups we also evaluated an acoustic keyword spotting system (AKWS) and a system searching for phone strings in lattices (WFSTlat). The core of this thesis is the development, analysis and improvement of the WFST QbE STD system, which after the improvements, achieved similar performance to the DTW system in language-dependent setups.
|
|
|
Query-by-Example Keyword Spotting
Skácel, Miroslav ; Hannemann, Mirko (oponent) ; Szőke, Igor (vedoucí práce)
The aim of the thesis is to get acquainted with modern approach of keyword spotting and spoken term detection in speech data. The bases of keyword spotting are described at first. The data representation used for experiments and evaluation are introduced. Keyword spotting methods where query is provided as an audio example (Query-by-Example) are presented. The scoring metrics are described and experiments follow. The results are discussed. Further, modern approaches of keyword spotting are suggested and implemented. The system with new techniques is evaluated and the discussion of results achieved follows. The conclusions are drawn and the discussion of future directions of development is held. The Appendix contains user manual for using implemented system.
|
|
|
Hledání akustických vzorů v řečových datech bez rozpoznávání
Skácel, Miroslav ; Fapšo, Michal (oponent) ; Černocký, Jan (vedoucí práce)
Tato práce se zabývá metodami vyhledávání slov, slovních frází a delších úseků v rozsáhlých řečových datech bez předchozích znalostí těchto dat. V úvodu je seznámení s danou problematikou a principy moderních metod pro vyhledávání opakujících se objektů. Dále je popsána reprezentace a segmentace vstupních dat, techniky pro vyhledání objektu v mluveném projevu a popis modelování nalezených objektů. Následně je popsána metoda pro vyhledávání objektů podle předem defi novaného vzoru. V dalším kroku jsou defi nována data pro experimenty, ve kterých byly použity metody pro detekci mluvených výrazů podle vzoru. Následuje popis systémových požadavků. V závěru je zhodnocení práce a návrhy na další vývoj.
|
|
|
Query-by-Example Spoken Term Detection
Fapšo, Michal ; Matoušek, Jindřich (oponent) ; Metze, Florian (oponent) ; Černocký, Jan (vedoucí práce)
This thesis investigates query-by-example (QbE) spoken term detection (STD). Queries are entered in their spoken form and searched for in a pool of recorded spoken utterances, providing a list of detections with their scores and timing. We describe, analyze and compare three different approaches to QbE STD, in various language-dependent and language-independent setups with diverse audio conditions, searching for a single example and five examples per query. For our experiments we used Czech, Hungarian, English and Levantine data and for each of the languages we trained a 3-state phone posterior estimator. This gave us 16 possible combinations of the evaluation language and the language of the posterior estimator, out of which 4 combinations were language-dependent and 12 were language-independent. All QbE systems were evaluated on the same data and the same features, using the metrics: non-pooled Figure-of-Merit and our proposed utterrance-normalized non-pooled Figure-of-Merit, which provided us with relevant data for the comparison of these QbE approaches and for gaining a better insight into their behavior. QbE approaches presented in this work are: sequential statistical modeling (GMM/HMM), template matching of features (DTW) and matching of phone lattices (WFST). To compare the performance of QbE approaches with the common query-by-text STD systems, for language-dependent setups we also evaluated an acoustic keyword spotting system (AKWS) and a system searching for phone strings in lattices (WFSTlat). The core of this thesis is the development, analysis and improvement of the WFST QbE STD system, which after the improvements, achieved similar performance to the DTW system in language-dependent setups.
|
|
|
Hledání akustických vzorů v řečových datech bez rozpoznávání
Skácel, Miroslav ; Fapšo, Michal (oponent) ; Černocký, Jan (vedoucí práce)
Tato práce se zabývá metodami vyhledávání slov, slovních frází a delších úseků v rozsáhlých řečových datech bez předchozích znalostí těchto dat. V úvodu je seznámení s danou problematikou a principy moderních metod pro vyhledávání opakujících se objektů. Dále je popsána reprezentace a segmentace vstupních dat, techniky pro vyhledání objektu v mluveném projevu a popis modelování nalezených objektů. Následně je popsána metoda pro vyhledávání objektů podle předem defi novaného vzoru. V dalším kroku jsou defi nována data pro experimenty, ve kterých byly použity metody pro detekci mluvených výrazů podle vzoru. Následuje popis systémových požadavků. V závěru je zhodnocení práce a návrhy na další vývoj.
|
|
|
Query-by-Example Keyword Spotting
Skácel, Miroslav ; Hannemann, Mirko (oponent) ; Szőke, Igor (vedoucí práce)
The aim of the thesis is to get acquainted with modern approach of keyword spotting and spoken term detection in speech data. The bases of keyword spotting are described at first. The data representation used for experiments and evaluation are introduced. Keyword spotting methods where query is provided as an audio example (Query-by-Example) are presented. The scoring metrics are described and experiments follow. The results are discussed. Further, modern approaches of keyword spotting are suggested and implemented. The system with new techniques is evaluated and the discussion of results achieved follows. The conclusions are drawn and the discussion of future directions of development is held. The Appendix contains user manual for using implemented system.
|
host ::
RSS.