A Computational Approach for Creating Orchestral Accompaniments from Piano Concerto Recordings (en)

* Presenting author
Day / Time: 08.03.2023, 10:20-10:40
Room: Saal Y8
Typ: Vortrag (strukturierte Sitzung)
Abstract: The piano concerto, composed for a pianist accompanied by an orchestra, is a genre of great importance in Western classical music. Even though most pianists practice piano concertos (also as an essential part of their piano education) in their careers, only first-class pianists have the opportunity to actually play with an orchestra. In this contribution, we propose a computational pipeline that allows pianists of any level to create their own mixes with an orchestral accompaniment coming from an existing recording. In particular, this pipeline consists of four components using techniques from music information retrieval (MIR). First, starting with a complete piano concerto recording, we apply data-driven source separation techniques to separate the piano and the orchestra. Second, we alleviate separation artifacts (e.g., musical noise) in a post-processing step. Third, we use music synchronization techniques to temporally align the separated orchestral track with the pianist's own recording. Finally, we apply time-scale modification to warp the orchestral track and create the final mix. While introducing a novel dataset used for training and testing our overall procedure, we discuss the various MIR techniques involved and report on first experimental results.


strpos(): Passing null to parameter #1 ($haystack) of type string is deprecated

/kunden/431424_13355/webseiten/subdomains/, line 35
32.              */
33.             if (headers_sent()) {
34.                 $encoding = false;
35.             } elseif (strpos($HTTP_ACCEPT_ENCODING, 'x-gzip') !== false) {
36.                 $encoding = 'x-gzip';
37.             } elseif (strpos($HTTP_ACCEPT_ENCODING, 'gzip') !== false) {
38.                 $encoding = 'gzip';