|
Sound Source Separation Demo This is a demonstration of our real-time sound source separation algorithm based on our Azimuth Discrimination and Resynthesis algorithm (ADRess 2004). Typical applications of the technology include, Demixing/Remixing and Upmixing, Noise Reduction and Source Intelligibility Enhancement and. The algorithm is also a useful pre-process for typical feature extraction algorithms. These demonstrations show the application operating in real-time complete with visual and auditory cues to help the user make the correct parameter selection. Example 1 shows the separation of a real recording of a jazz quartet. Example 2 shows the application separating a speech mixture. Both examples are in .avi format. *This research is the property of the Dublin Institute of Technology and is protected by patent. |
VIDEO DEMO 1 VIDEO DEMO 2
|
|
Time and Pitch Scale Demo This is a demonstration of our work on time and pitch scale modification. This method is based on a phase locked phase vocoder approach. The addition of real-time transient detection in this application allows for an almost “artefact free” time scaling for most musical applications. The technique preserves transients very accurately allowing for very high quality time scaling in the region of 50% to 200%. Independent pitch shifting is also achieved in the implementation. The demonstration shows the responsiveness and flexibility of the approach. Example 1 and 2 are musical examples in .avi format. *This research is the property of the Dublin Institute of Technology. These technologies are currently available for licence. Contact via dan.barry@dit.ie |
VIDEO DEMO 1 VIDEO DEMO 2
|
|
Real-time Beat Editing and Mashup This is a demonstration of our real-time beat editing and mashup technology. It uses several of our algorithms to automatically calcualte the tempo, beat and barline positions from raw audio. This is illustrated in the video with the virtual metronome function which can be heard sounding a tone over each beat and bar line. These positions can then be used to control playback allowing perfectly beat synchronous loops which can be played in any order thus allowing the user to create non-linear playback effects resulting in real-time mashups of the audio. This is totally non-destructive since it is a real-time process on the audio stream. This is then illustrated by applying time and pitch scaling to the mashup stream. *This research is the property of the Dublin Institute of Technology and is protected by patent. |
|
Use of this technology or demonstrations is not permitted without prior consent of DIT