When interacting with the Smart environment using speech interfaces, one of the important aspects is to assess the quality of the pronunciation of phrases. Therefore, obtaining objective quantitative estimates of the proximity to the initial standard as one of the quality measures is relevant. The paper proposes a new method for assessing the quality of phrases pronunciation based on a user authentication approach using deep learning of neural networks. This approach can be used to assess the proximity of the presented sample within the Smart-environment in relation to the initial standard. Such an assessment can be useful both in determining the quality of pronunciation of repeated phrases in relation to the reference one (for example, for assessing the channel used when interacting with the Smart environment or for assessing the quality of the speaker's speech to identify qualitative changes related to his condition or health), so and directly during the procedure for confirming the identity of the speaker. A significant correlation of a new approach for these tasks in comparison with the existing ones based on speech recognition (for quality assessment tasks) is shown.
25 August 2019 in International Journal of Advanced Trends in Computer Science and Engineering
The article considers an approach to the problem of assessing the quality of speech during speech rehabilitation as a classification problem. For this, a classifier is built on the basis of an LSTM neural network for dividing speech signals into two classes: before the operation and immediately after. At the same time, speech before the operation is the standard to which it is necessary to approach in the process of rehabilitation. The metric of belonging of the evaluated signal to the reference class acts as an assessment of speech. An experimental assessment of rehabilitation sessions and a comparison of the resulting assessments with expert assessments of phrasal intelligibility were carried out.