When evaluating the effectiveness of speaker identification systems, an important role is played not only by the correct and properly performed extraction of speech signal features and the selection of appropriate speaker identification algorithms, but also by the correct acquisition and recording of voice signals with minimal distortion for later comparison with a reference pattern. This paper describes the audio path between the speaker’s mouth and the decision-making system, presents classifications of sources of noise and distortion accompanying the sound recording process, and classifies methods and algorithms for improving speech signal quality.Modern speaker identification and verification systems are based on increasingly complex decision-making processes. However, regardless of the methods or algorithms used to identify the speaker, it is extremely important to provide (…)
The article contains 25,266 characters.
Source: Logistyka Magazine
Methods for improving sound quality in speaker identification systems – PDF file for download