![]() ![]() Human speech recognition performance in realistic spatial listening conditions is affected by many non-linearly interacting factors, even when only the acoustic modality is considered. Its aim is to demonstrate the potential of future speech perception models in a detailed example, and to motivate thinking about suitable evaluation criteria for such models. However, its aim is not to compare or discuss the prediction accuracy of existing models that predict speech intelligibility. The focus lies on presenting a concrete example application with exemplary data using existing model components in the context of speech perception with impaired hearing. In this contribution, as a first step towards a scenario where predictions of human speech recognition performance are accurate and inexpensive, an accessible visual representation of extensive modeling data is presented and investigated. But huge amounts of data are not necessarily useful by themselves, which raises the question which data would be useful to predict in the first place. In any case, it will mean to predict a lot of data. Natural applications could be: hearing aid fitting/evaluation, speech-perception aware acoustic room/space design, or virtual assessments of hearing-related technology with respect to speech perception. It also adds a problem-centric perspective to the often rather model-centric perspectives on speech intelligibility, because it asks for concrete applications. “ What could one do with a speech perception model that accurately predicts the individual speech recognition performance of a listener in any listening situation?” While this still remains a philosophical question, with the advances in automatic speech recognition (ASR) and especially speech recognition modeling, it will become more tangible in the future. This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( ), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Ultimately, it may serve as a tool to use validated prediction models in the design of spaces and devices which take speech communication into account. The proposed representation was found to be a suitable target to compare and validate modeling approaches in ecologically relevant contexts, and should help to explore possible applications of future speech recognition models. Using the spatial speech recognition maps to explore this data set demonstrated the suitability of the approach to observe possibly relevant listener behavior. An exemplary modeling toolchain, that is, a combination of an acoustic model, a hearing device model, and a listener model, was used to generate a data set for demonstration purposes. In this contribution, an interactive, that is, user manipulable, representation of speech recognition performance is proposed and investigated by means of a concrete example, which focuses on the listener’s head orientation and the spatial dimensions – in particular width and depth – of an acoustic scene. This raises an interesting question: What could we do if we had a perfect speech intelligibility model? In a first step, means to explore and interpret the predicted outcomes of large numbers of speech recognition experiments would be helpful, and large amounts of data demand an accessible, that is, easily comprehensible, representation. individual hearing aid fitting, they can already be performed. While such predictions may still not be sufficiently accurate for serious applications, such as, e.g. ![]() multi-source dynamic, acoustic environments. Prediction models come closer to considering all required factors simultaneously to predict the individual speech recognition performance in complex, that is, e.g. * Corresponding author: their everyday life, the speech recognition performance of human listeners is influenced by diverse factors, such as the acoustic environment, the talker and listener positions, possibly impaired hearing, and optional hearing devices. Medizinische Physik and Cluster of Excellence Hearing4all, Universität Oldenburg, 26111 Oldenburg, Germany
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |