The problem of video framerate and audio/video synchronization in audio-visual speech recognition is considered. The visual features are added to the acoustic parameters in order to improve the accuracy of speech recognition in noisy conditions. The Mel-Frequency Cepstral Coefficients are used on the acoustic side whereas Active Appearance Model features are extracted from the image. The feature fusion approach is employed. The initial video framerate equals 100 frames per second. The test signals were recorded with a specialized hardware for synchronous registration of audio and video data. In a practical implementation, however, it is difficult to achieve a high rate of images per second and maintain the precise audio/video synchronization. Therefore, in this work it is assessed, how the lowered framerate and lack of synchronization between audio and video data impairs the performance of the recognition engine. The lowered video framerate is enforced by downsampling the visual data. The lack of synchronization is simulated programmatically in the feature fusion process. The experiments are conducted employing the HTK engine (Hidden Markov Toolkit). Word Error Rate, correctness and accuracy measures are considered, while a small dictionary of 11 words (numerals) is employed.
Authors
Additional information
- Category
- Publikacja monograficzna
- Type
- rozdział, artykuł w książce - dziele zbiorowym /podręczniku w języku o zasięgu międzynarodowym
- Language
- angielski
- Publication year
- 2014
