Improving the efficiency of multi-channel voice recognition system

• Main Author: Александр Анатольевич Штепа
• Format: Article
• Language: English
• Published: PC Technology Center 2015-09-01
• Series: Tehnologìčnij Audit ta Rezervi Virobnictva
• Subjects: voice recognition; adaptive compensation; proximity measure
• Online Access: http://journals.uran.ua/tarp/article/view/51796
• ISSN: 2226-3780; 2312-8372

The use of methods of noisy signal treatment in voice recognition is discussed and some of the research results in this area are given. The main aim of the study is improving the effectiveness of recognition system of voice commands in a difficult acoustic environment by improving the signal / noise ratio due to the use of the spatial separation of signals using multiple directional microphones and digital signal processing on the basis of adaptive interference cancellation. The use of modern methods of language voice recognition together with the adaptive compensation method for processing a noisy signal can improve the accuracy of voice recognition. The algorithmic and structural approaches to solving the problems of sound processing, resulting in difficult acoustic conditions for further recognition of voice commands are discussed in the article. The presented method allows increasing the accuracy of the definition of basic and auxiliary channels of multi-channel voice recognition system to increase the efficiency of adaptive compensation. The method and the algorithm are designed to automatically determine the most appropriate channel for base and support channels in accordance with the method of adaptive interference cancellation. We propose to use proximity measure between the received signal and the standard obtained at command recognition on the basis of non-linear time alignment as a criterion for determining the base channel. The research results can be applied to voice recognition in the voice control systems of equipment and vehicles.