Identification of Audio Coding Artifacts Generated Due to Bandwidth Extension Schemes

Identification of audio coding artifacts has a wide range of applications, including encoder design, audio post-processing, and quality assessment. In this paper, we focus on identifying audio artifacts associated with (semi-)parametric audio coding, particularly in the context of audio bandwidth extension (BWE) schemes. The application of BWE can lead to artifacts in the high-frequency target area that depend on the crossover frequency and the characteristics of the signal. We aim to identify on a frame-by-frame basis two common types of artifacts: Tonality Mismatch (TM) and Unmasked Noise (UN). To achieve this, we introduce a novel method for detecting these artifact types, incorporating two key components into our model: a spectral flatness measure and harmonic analysis of the signal.