A Decade of Encouraging Speech Processing “Outside of the Box”—A Foreword

Abstract

A perspective on this year’s Non-Linear Speech Processing effort is given focussing on alternative approaches. In fact, it was marked by a strong presence of the “younger” field of paralinguistic speech analysis. Beyond the more mature recognition of emotion, a significant interest in health and care applications can be noticed. This includes Alzheimer’s disease, cognitive load level, depressive disorders, and Parkinson’s disease, and reaches to language impairment, general pathology as well as speech interfaces for the elderly. In fact, even speaker identification has seen a study on the influence of emotionally converted speech. Related to speech analysis, the front-end side along the usual chain of speech processing was strongly represented. Backend machine learning includes extreme learning machines, unsupervised clustering, and deep representations.