| Katharine Mary Scott |
Audio source separation is an application of digital signal processing that targets the extraction of individual audio sources from composite audio signals. The ability to accurately separate audio sources is crucial for enhancing clarity and quality in environments where overlapping sounds present significant challenges. This study illustrates the application of Empirical Mode Decomposition (EMD) and Hilbert-Huang Transform (HHT) within SAS/IML to dissect a composite audio signal into its constituent sources. Starting with a merged audio file containing signals from different sources, the process begins with an individual analysis of baseline audio files to establish a baseline frequency distribution for each source. Utilizing EMD, the composite signal is decomposed into Intrinsic Mode Functions (IMFs), facilitating the isolation of distinct audio frequencies attributed to each source. The HHT is employed subsequently to scrutinize these IMFs in the frequency domain, ensuring the accuracy of separation. This approach is exemplified by successful separation of the distinct audio frequencies, where specific IMFs demarcate the distinct sources. The methodology concludes with the reconstruction of separated audio files, confirming the practical effectiveness through auditory verification. This presentation illustrates the capabilities of SAS/IML in handling complex audio processing tasks and provides a model for similar challenges in audio source separation. The applications of this technique extend to various fields requiring audio segmentation, such as multimedia editing, forensic analysis, and automated transcription services. |