Audio feedback on metal panels and emotion detection for sound classification in low-hierarchy performance spaces

Abstract

Social inequality in both the classical music production and consumption practices is undoubtedly at the base of the lack of involvement of the so-called new audiences in the contemporary music scene and it has been observed that timbre-based approaches to synthesis tends to favour listening and interaction when people have no music background. The present paper investigates the use of audio and electromagnetic feedback for the stimulation of vibration in metal panels and metallic musical instruments within the frame of interactive sound installations and compositions to pursue the exploration of less hierarchical structures within the performance space. A variety of feedback algorithms implemented for those works will be described along with the machine learning techniques used to expand the human-machine relationship via web-applications, which allow for remote interaction of the audience. Classification of sound according to its emotional content was used because non-musicians tend to relate to music mainly through their emotional response to it. Sound analysis and re-synthesis enabled the production of control data to steer AC devices as well as regulate the activation of the audio feedback algorithms. Given its non-linearity and its reluctance to control, feedback turned out to be particularly successful in the generation of a vast variety of timbre-based audio material and agency, thus laying the ground for free improvisation and interaction within the performance space.