A scalable architecture for multilingual speech recognition on embedded devices

Abstract

Please cite this article as: Raab, M., Gruhn, R., Nöth, E., A scalable architecture for multilingual speech recognition on embedded devices, Speech Communication (2010), doi: 10.1016/j.specom. 2010.07.007 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. AbstractIn-car infotainment and navigation devices are typical examples where speech based interfaces are successfully applied. While classical applications are monolingual, such as voice commands or monolingual destination input, the trend goes towards multilingual applications. Examples are music player control or multilingual destination input. As soon as more languages are considered the training and decoding complexity of the speech recognizer increases. For large multilingual systems, some kind of parameter tying is needed to keep the decoding task feasible on embedded systems with limited resources. A traditional technique for this is to use a semi-continuous Hidden Markov Model as the acoustic model. The monolingual codebook on which such a system relies is not appropriate for multilingual recognition. We introduce Multilingual Weighted Codebooks that give good results with low decoding complexity. These codebooks depend on the actual language combination and increase the training complexity. Therefore an algorithm is needed that can reduce the training complexity. Our first proposal are mathematically motivated projections between Hidden Markov Models defined in Gaussian spaces. Although theoretically optimal, these projections were difficult to employ directly in speech decoders. We found approximated projections to be most effective for practical application, giving good performance without requiring major modifications to the common speech recognizer architecture. With a combination of the Multilingual Weighted Codebooks and Gaussian Mixture Model projections we create an efficient and scalable architecture for non-native speech recognition. Our new architecture offers a solution to the combinatoric problems of training and decoding for multiple languages. It builds new multilingual systems in only 0.002% of the time of a traditional HMM training, and achieves comparable performance on foreign languages.

Development of robust electrooculography (EOG)-based human-computer interface controlled by eight-directional eye movements

“…Currently, many research studies are underway into means of enabling the disabled and elderly to communicate effectively with machine or computer. Depending on the users' capabilities, different types of interface have been proposed, such as speech recognition based on both voice (Raab et al, 2011) and surface electromyography (Fraiwan et al, 2011), lip movement control system (Shaikh et al, 2011), visionbased multiple gestures (Reale et al, 2011), sip-and-puff *Corresponding author. E-mail: angkoon.p@hotmail.com.…”

Modelo Acústico y de Lenguaje del Idioma Español para el dialecto Cucuteño, Orientado al Reconocimiento Automático del Habla

“…Luego de realizar una breve descripción de estos sistemas de procesamiento de voz, es importante entender la manera como se desarrollan y entrenan los diferentes elementos que componen el reconocedor. En este punto es importante aclarar que el RAH está desarrollado bajo el software de código abierto CMU Sphinx [17] el cual es instalado y ejecutado en el ordenador de placa reducida Raspberry PI B+, de sistema operativo Raspbian [18]. Como se mencionó anteriormente los modelos del lenguaje y acústico se desarrollan para satisfacer las necesidades de un RAH de múltiples hablantes y de un sistema de control por comandos.…”

EOG Based Text and Voice Controlled Remote Interpreter for Quadriplegic Patients

“…Many research projects are currently looking into ways to help the elderly and disabled communicate effectively with machines or computers. Various interface types have been proposed depending on the users' capabilities, including; Speech recognition based on both voice [2], Surface electromyography [3], lip movement control system [4], infrared and ultrasonic non-contact head controllers [5], multifunction myoelectric control system (MMCS) [6], brain-computer interface (BCI) [7].…”