Selen Hande Kabil

Official source

À propos de ce résultat

Cette page est générée automatiquement et peut contenir des informations qui ne sont pas correctes, complètes, à jour ou pertinentes par rapport à votre recherche. Il en va de même pour toutes les autres pages de ce site. Veillez à vérifier les informations auprès des sources officielles de l'EPFL.

Unités associées

Chargement

Cours enseignés par cette personne

Chargement

Domaines de recherche associés

Chargement

Publications associées

Chargement

Personnes menant des recherches similaires

Chargement

Sparse Autoencoders for Speech Modeling and Recognition

Selen Hande Kabil

Speech recognition-based applications upon the advancements in artificial intelligence play an essential role to transform most aspects of modern life. However, speech recognition in real-life conditions (e.g., in the presence of overlapping speech, varying speaker characteristics) remains to be a challenge. The current state of the research to achieve robust speech recognition mostly depends on building systems driven by complex deep neural networks. Nonetheless, speech production process enables low-dimensional subspaces which can carry class-specific information in speech. In this thesis, we investigate the exploitation of this low-dimensional multi-subspace structure of speech towards the goal of improving acoustic modeling for automatic speech recognition (ASR).This thesis mainly focuses on the sparse autoencoders for sparse modeling of speech, starting from their often-overlooked connection with sparse coding. We hypothesize that whenever speech signal is represented in a high-dimensional feature space, the true class information (regarding the speech content) is embedded in low-dimensional subspaces. The analysis on the high-dimensional sparse speech representations obtained from the sparse autoencoders demonstrates their prominent capability of modeling the underlying (e.g., sub-phonetic) components of speech. When used for recognition, the representations from sparse autoencoders yield performance improvements. Finally, we repurpose the aforementioned sparse autoencoders for pathological speech recognition task in transfer learning framework.In this context, the contribution of this thesis is twofold: (i) in speech modeling, proposing the use of sparse autoencoders as a novel way of sparse modeling for extracting the class-specific low-dimensional subspaces in speech features, and (ii) in speech recognition, demonstrating the effectiveness of these autoencoders in the state-of-the-art ASR frameworks towards the goal of improving robust ASR, in particular on far-field speech from AMI and pathological speech from UA-Speech datasets.

EPFL2023

Chargement

In Bourlard and Kamp (Biol Cybern 59(4):291-294, 1998), it was theoretically proven that autoencoders (AE) with single hidden layer (previously called "auto-associative multilayer perceptrons") were, in the best case, implementing singular value decomposition (SVD) Golub and Reinsch (Linear algebra, Singular value decomposition and least squares solutions, pp 134-151. Springer, 1971), equivalent to principal component analysis (PCA) Hotelling (Educ Psychol 24(6/7):417-441, 1993); Jolliffe (Principal component analysis, springer series in statistics, 2nd edn. Springer, New York ). That is, AE are able to derive the eigenvalues that represent the amount of variance covered by each component even with the presence of the nonlinear function (sigmoid-like, or any other nonlinear functions) present on their hidden units. Today, with the renewed interest in "deep neural networks" (DNN), multiple types of (deep) AE are being investigated as an alternative to manifold learning Cayton (Univ California San Diego Tech Rep 12(1-17):1, 2005) for conducting nonlinear feature extraction or fusion, each with its own specific (expected) properties. Many of those AE are currently being developed as powerful, nonlinear encoder-decoder models, or used to generate reduced and discriminant feature sets that are more amenable to different modeling and classification tasks. In this paper, we start by recalling and further clarifying the main conclusions of Bourlard and Kamp (Biol Cybern 59(4):291-294, 1998), supporting them by extensive empirical evidences, which were not possible to be provided previously (in 1988), due to the dataset and processing limitations. Upon full understanding of the underlying mechanisms, we show that it remains hard (although feasible) to go beyond the state-of-the-art PCA/SVD techniques for auto-association. Finally, we present a brief overview on different autoencoder models that are mainly in use today and discuss their rationale, relations and application areas.

SPRINGER2022

Chargement

From Undercomplete to Sparse Overcomplete Autoencoders to Improve LF-MMI Speech Recognition

Hervé Bourlard, Selen Hande Kabil

Starting from a strong Lattice-Free Maximum Mutual Information (LF-MMI) baseline system, we explore different autoencoder configurations to enhance Mel-Frequency Cepstral Coefficients (MFCC) features. Autoencoders are expected to generate new MFCC features that can be used in our LF-MMI based baseline system (with or without retraining) towards speech recognition improvements. Starting from shallow undercomplete autoencoders, and their known equivalence with Principal Component Analysis (PCA), we go to deeper or sparser architectures. In the spirit of kernel-based learning methods, we explore alternatives where the autoencoder first goes overcomplete (i.e., expand the representation space) in a nonlinear way, and then we restrict the autoencoder by means of a sequent bottleneck layer. Finally, as a third solution, we use sparse overcomplete autoencoders where a sparsity constraint is imposed on the higher-dimensional encoding layer. Experimental results are provided on the Augmented Multiparty Interaction (AMI) dataset, where we show that all aforementioned architectures improve speech recognition performance.

ISCA-INT SPEECH COMMUNICATION ASSOC2022

Chargement