Abstract

In this paper, we address the challenges faced when combining noise cancellation and automatic speech recognition models. When these models are combined directly, the performance of word recognition often suffers because the distribution of input data changes. To overcome this limitation, we propose a novel method for combining these models, which enhances the ability of the speech recognition model to perform well in noisy environments. The key feature of the proposed method is the introduction of a mechanism to control the aggressiveness of noise reduction. This mechanism enables us to customize the noise reduction process according to the specific requirements of the ASR model, without necessitating any retraining. This advantage makes our method applicable to any ASR model, facilitating its implementation in practical scenarios.

‹ 1 / ? › − 100% + Open in new tab Download Cite

Loading the full text…