The most important challenges in AVSR and the focus of most research are the features that are extracted, and when combined give better results. The other challenge is the resulted feature here of nature are large in size, then prefers here to reduce the features by use of an appropriate way to reduce these data with ensure have their properties after downsizing. The System that is presented in this research is for recognition a group of Arabic words voices, from one to ten words. In the acoustic parts the features were extracted of coefficients MFCC, LPC,FFT to be determine which type of these features is efficient in AVSR .All these types of feature are showed efficient results but MFCC is the best. The visual features are calculated of DCT matrix, and the features are extracted by applying the zigzag scan. In the reduction features stage, several methods of data reducing have been implemented; they are LDA, PCA and SVD. Each method are applied to the data separately. The KNN models are used in the stage of recognition, where the testing is implemented on dependent and independent database of words from one to ten. The final results that obtained are efficient and encouraging.

Emerging broadband communication systems promise a future of multimedia telephony, e.g. the addition of visual information to telephone conversations. It is useful to consider the problem of generating the critical information useful for speechreading, based on existing narrowband communications systems used for speech. This paper focuses on the problem of synthesizing visual articulatory movements given the acoustic speech signal. In this application, the acoustic speech signal is analyzed and the corresponding articulatory movements are synthesized for speechreading. This paper describes a hidden Markov model (HMM)-based visual speech synthesizer. The key elements in the application of HMMs to this problem are the decomposition of the overall modeling task into key stages and the judicious determination of the observation vector's components for each stage. The main contribution of this paper is a novel correlation HMM model that is able to integrate independently trained acoustic and visual HMMs for speech-to-visual synthesis. This model allows increased flexibility in choosing model topologies for the acoustic and visual HMMs. Moreover the propose model reduces the amount of training data compared to early integration modeling techniques. Results from objective experiments analysis show that the propose approach can reduce time alignment errors by 37.4% compared to conventional temporal scaling method. Furthermore, subjective results indicated that the purpose model can increase speech understanding.

No abstract

No abstract

No abstract