Rasmus Dall

Reactive Statistical Mapping: Towards the Sketching of Performative Control with Data

This paper presents the results of our participation to the ninth eNTERFACE workshop on multimodal user interfaces. Our target for this workshop was to bring some technologies currently used in speech recognition and synthesis to a new level, i.e. being the core of a new HMM-based mapping system. The idea of statistical mapping has been investigated, more precisely how to use Gaussian Mixture Models and Hidden Markov Models for realtime and reactive generation of new trajectories from inputted labels and for realtime regression in a continuous-to-continuous use case. As a result, we have developed several proofs of concept, including an incremental speech synthesiser, a software for exploring stylistic spaces for gait and facial motion in realtime, a reactive audiovisual laughter and a prototype demonstrating the realtime reconstruction of lower body gait motion strictly from upper body motion, with conservation of the stylistic properties. This project has been the opportunity to formalise HMM-based mapping, integrate various of these innovations into the Mage library and explore the development of a realtime gesture recognition tool.

Redefining the Linguistic Context Feature Set for HMM and DNN TTS Through Position and Parsing.

In this paper we present an investigation of a number of alternative linguistic feature context sets for HMM and DNN textto-speech synthesis. The representation of positional values is explored through two alternatives to the standard set of absolute values, namely relational and categorical values. In a preference test the categorical representation was found to be preferred for both HMM and DNN synthesis. Subsequently, features based on probabilistic context free grammar and dependency parsing are presented. These features represent the phrase level relations between words in the sentences, and in a preference evaluation it was found that these features all improved upon the base set, with a combination of both parsing methods best overall. As the features primarily affected the F0 prediction, this illustrates the potential of syntactic structure to improve prosody in TTS.

Synthesising Filled Pauses: Representation and Datamixing.

Filled pauses occur frequently in spontaneous human speech, yet modern text-to-speech synthesis systems rarely model these disfluencies overtly, and consequently they do not output convincing synthetic filled pauses. This paper presents a text-to-speech system that is specifically designed to model these particular disfluencies more efffectively. A preparatory investigation shows that a synthetic voice trained exclusively on spontaneous speech is perceived to be inferior in quality to a voice trained entirely on read speech, even though the latter does not handle filled pauses well. This motivates an investigation into the phonetic representation of filled pauses which show that, in a preference test, the use of a distinct phone for filled pauses is preferred over the standard /V/ phone and the alternative /@/ phone. In addition, we present a variety of data-mixing techniques to combine the strengths of standard synthesis systems trained on read speech corpora with the supplementary advantages offered by systems trained on spontaneous speech. In a MUSHRA-style test, it is found that the best overall quality is obtained by combining the two types of corpora using a source marking technique. Specifically, general speech is synthesised with a standard mark, while filled pauses are synthesised with a spontaneous mark, which has the added benefit of also producing filled pauses that are comparatively well synthesised.