Multi-style learning with denoising autoencoders for acoustic modeling in the internet of things (IoT). (November 2017)

• Record Type: Journal Article
• Authors: Lin, Payton; Lyu, Dau-Cheng; Chen, Fei; Wang, Syu-Siang; Tsao, Yu
• Is Part Of: Computer speech & language. Volume 46(2017)
• Page Start: 481
• Page End: 495
• Publication Date: 2017-11
• Subjects: Deep learning -- Deep neural networks -- Multi-style training -- Deep denoising autoencoders -- Mixed training -- Representation learning -- Data combination -- Data synthesis -- Noise injection theory -- Feature compensation -- Automatic speech recognition -- Internet of things (IoT); Speech processing systems -- Periodicals; Automatic speech recognition -- Periodicals; Computers -- Periodicals; Linguistics -- Periodicals; Speech-Language Pathology -- Periodicals; Traitement automatique de la parole -- Périodiques; Reconnaissance automatique de la parole -- Périodiques; Automatic speech recognition; Speech processing systems; Electronic journals; Periodicals; 006.454
• DOI: 10.1016/j.csl.2017.02.001
• Languages: English
• ISSNs: 0885-2308

Abstract

Highlights: Multi-style Learning (Multi-style training+Deep Learning) relies on multiple layers of nonlinear data transformations to synthesize additional examples of representative variations. Rather than artificially creating data samples, a deep denoising autoencoder (DAE) is instead utilized to extract and organize the most discriminative information in a training set. Multi-style learning combines the DAE-synthesized data with the original data to make class boundaries less sensitive to corruptions by enforcing the back-end model to emphasize on the most discriminative patterns. The proposed data-mixed deep neural networks (DNNs) consistently improved the performance without requiring any preprocessing on the test sets. Abstract: We propose a multi-style learning ( multi-style training + deep learning ) procedure that relies on deep denoising autoencoders (DAEs) to extract and organize the most discriminative information in a training database. Traditionally, multi-style training procedures require either collecting or artificially creating data samples (e.g., by noise injection or data combination) and training a deep neural network (DNN) with all of these different conditions. To expand the applicability of deep learning, the present study instead adopts a DAE to augment the original training set. First, a DAE is utilized to synthesize data that captures useful structure in the input distribution. Next, this synthetic data is combined and mixed within the original training set to exploit the powerful capabilities of DNN classifiers to learn the complex decision boundaries in heterogeneous conditions. By assigning a DAE to synthesize additional examples of representative variations, multi-style learning makes class boundaries less sensitive to corruptions by enforcing back-end DNNs to emphasize on the most discriminative patterns. Moreover, this deep learning technique mitigates the cost and time of data collection and is easy to incorporate into the internet of things (IoT). Results showed these data-mixed DNNs provided consistent performance improvements without even requiring any preprocessing on the test sets.