Sonic Gestures and Rhythmic Interaction between the Human and the Computer

Aalto University, P.O. Box 11000, FI-00076 Aalto www.aalto.fi Author Antti Jylhä Name of the doctoral dissertation Sonic Gestures and Rhythmic Interaction between the Human and the Computer Publisher School of Electrical Engineering Unit Department of Signal Processing and Acoustics Series Aalto University publication series DOCTORAL DISSERTATIONS 32/2012 Field of research Acoustics and audio signal processing Manuscript submitted 15 December 2011 Manuscript revised 15 March 2012 Date of the defence 20 April 2012 Language English Monograph Article dissertation (summary + original articles) Abstract This thesis addresses the use of sonic gestures as input in human-computer interaction with a special applicability focus on rhythmic interactive systems and their design and evaluation. Sonic gestures are defined as human-generated sounding actions which convey information to a computational system. Examples of such gestures are impulsive sounding actions such as hand claps and finger snaps, sustained actions such as humming and blowing, and iterative actions, such as tapping a table to the beat of music. The use of sonic gestures at the interface requires analysis algorithms that are capable of tracking the desired information from an audio stream containing the human-generated sounds. In interactive systems, these algorithms must be capable of real-time processing. In this thesis, the focus is on percussive sonic gestures, which can be seen to be analogous to the sounds of percussive instruments. Therefore, it is reasonable to assume that the same tools that are applied for retrieving information from drums and percussion in music can be deployed for sonic gesture analysis. This work presents algorithms for the classification of different percussive sounds, such as different types of hand claps. To demonstrate the use of sonic gestures, a hand clap interface capable of recognizing different hand clap types and extracting continuous information, such as the tempo, from a clapping sequence has been developed. This interface has been utilized in the development of various rhythmic prototype applications, most importantly a system called iPalmas, an interactive Flamenco rhythm tutor. The iPalmas system can produce realistic-sounding synthetic Flamenco hand clapping patterns to the user, listen to the clapping of the user, and give audiovisual feedback on the learning and performance. The iPalmas system was evaluated in a subjective experiment, resulting in qualitative and quantitative findings related to the system design, the human capabilities, and the interaction. In conjunction with this evaluation, a structured framework for evaluating this kind of systems has been proposed. Based on the evaluation results, the system has undergone iterative development of the audiovisual feedback elements. The main outcomes of the thesis are a novel definition of sonic gestures in human-computer interaction and a taxonomy of the information they can convey to computational systems and the interactive iPalmas system, resulting in several relevant findings that can be generalized in the design and evaluation of rhythmic interactive systems.This thesis addresses the use of sonic gestures as input in human-computer interaction with a special applicability focus on rhythmic interactive systems and their design and evaluation. Sonic gestures are defined as human-generated sounding actions which convey information to a computational system. Examples of such gestures are impulsive sounding actions such as hand claps and finger snaps, sustained actions such as humming and blowing, and iterative actions, such as tapping a table to the beat of music. The use of sonic gestures at the interface requires analysis algorithms that are capable of tracking the desired information from an audio stream containing the human-generated sounds. In interactive systems, these algorithms must be capable of real-time processing. In this thesis, the focus is on percussive sonic gestures, which can be seen to be analogous to the sounds of percussive instruments. Therefore, it is reasonable to assume that the same tools that are applied for retrieving information from drums and percussion in music can be deployed for sonic gesture analysis. This work presents algorithms for the classification of different percussive sounds, such as different types of hand claps. To demonstrate the use of sonic gestures, a hand clap interface capable of recognizing different hand clap types and extracting continuous information, such as the tempo, from a clapping sequence has been developed. This interface has been utilized in the development of various rhythmic prototype applications, most importantly a system called iPalmas, an interactive Flamenco rhythm tutor. The iPalmas system can produce realistic-sounding synthetic Flamenco hand clapping patterns to the user, listen to the clapping of the user, and give audiovisual feedback on the learning and performance. The iPalmas system was evaluated in a subjective experiment, resulting in qualitative and quantitative findings related to the system design, the human capabilities, and the interaction. In conjunction with this evaluation, a structured framework for evaluating this kind of systems has been proposed. Based on the evaluation results, the system has undergone iterative development of the audiovisual feedback elements. The main outcomes of the thesis are a novel definition of sonic gestures in human-computer interaction and a taxonomy of the information they can convey to computational systems and the interactive iPalmas system, resulting in several relevant findings that can be generalized in the design and evaluation of rhythmic interactive systems.