Performance Estimation of Noisy Speech Recognition Considering the Accuracy of Acoustic Models

To ensure a satisfactory QoE (Quality of Experience) and facilitate system design in speech recognition services, it is essential to establish a method that can be used to efficiently investigate recognition performance in different noise environments. Previously, we proposed a performance estimation method using the PESQ (Perceptual Evaluation of Speech Quality) as a measure of speech distortion. However, there is the problem that the accuracy of acoustic models used for speech recognition affects the relationship between the recognition performance and the distortion value. To solve this problem, we propose a novel performance estimation method considering the accuracy of acoustic models. Experimental results confirmed that the proposed method gives accurate estimates of the recognition performance for different sets of acoustic models, when using the recognition error rate for clean speech as a measure of the accuracy of acoustic models.