Taikichiro Mori Memorial Research Grant Report 2013

 

Influence on humanís emotions by means of Western Classical and Popular Music

 

Tatiana Endrjukaite,

PhD Student 1st year,

Graduate School of Media and Governance

 

Abstract:

It is well established that people respond emotionally to music, little is known about precisely what it is in the music that they are responding to. In this research we study dependency between intrinsic characteristics of tunes and emotions people experience while listening to those tunes. The feature of this research is that it explores ways of influencing on peopleís emotions through music. The objective of this research is to construct a system for determining the expected emotional effect of listening to a tune.

Introduction:

At present, there is no complete theory for automatically analysing music structure. However, there are encouraging researches on music signal processing and retrieval detecting the most frequently appearing component in a piece of music based on music structure analysis. Generally speaking, the more repetitions and similar phases there are in a piece of music, the easier it is for people to have affinity for it.

The main motivation of this research is to determine music characteristics mapping to emotions. Regarding that motivation, researches have made genre recognition analysis and music similarity analysis by detecting the most representative part of a piece of music. The most representative part in a piece has been defined as the most frequently repeated component in it such as repetitions and typical parts during tuneís performance. In particular, PCM acoustic data has been used. Repetitions have found by performing self-similarity calculations with Mel-Frequency Cepstral Coefficients. Tuneís internal homogeneity automatic identification by typical parts within tune have been detected by performing instantaneous frequency spectrum (IFS).

The relation between musical sounds and their influence on the listenerís emotion were studied by Hevner through experiments, which substantiated a hypothesis that music inherently carries emotional meaning. We want to find information extractable directly from music that, after appropriate processing, will help people to find new and unknown tunes they likely would like to listen to. The information we seek is mainly discussed in acoustic psychology and as such, its significance is an on-going research topic. Accordingly, this research should be based on music psychology and at the same time this research should contribute to musical research.

Study and Objectives:

Very large number of music pieces exist in the world today. People listen to different genres of music, such as classical, popular, jazz, blues and others. It was observed, that people usually listen to the same kind of music they prefer. For example, same artist usually performs his singles in the same manner, mood, and musical genre. To find a new tune we will like, we have to listen to this music piece to find out whether we will like it or not. As soon as various structures in music influence on us, we have to consider something more than just separate characteristics such as tonality, timbre, speed, or pitch. Repetitive parts in music pieces contain the most representative and significant information about the tune. Generally speaking, the more repetitions and similar phases there are in a piece of music, the easier it is for people to have affinity for it. For example, in the popular music style choruses are repeated several times in the tune. That makes it possible to easily remember the tune and recall it from the memory when we want to listen something similar.

There are two main task of this reseach:

1)    To find and presents similarity of music tunes by means of repeated parts in it. That will help people to find music pieces they may be even do not know, but most probably will like. Such tunes can have different tonality and may be even have different genre, but sound very similar to tunes the person prefers to listen.

2)    Detect music internal homogeneity. This task contributes to automatic homogeneity analysis of the music signal for selecting most typical parts within tunes.

The ultimate goal of this research is to construct a system for determining the expected emotional effect of listening to a tune and to determine mapping between music characteristics and emotions.

 

Constructed Music System:

The system of music processing to detect tuneís emotion is outlined in figure 1, where white boxes show data and blue boxes show processing steps. Acoustic data is used as an input. Output is a detected tuneís emotion. Tunes database store information about every tune and its corresponding emotion that was determined statistically. Tune itself is presented in the database by the tune descriptor, which contains three significant physical parameters of the tunes. Emotion for a tune is stored as a point on the emotional plane. This makes it possible to aggregate emotions from multiple tunes into a one single estimation.

 

Figure 1. System for determining the expected emotional effect of listening to a tune.

 

Results of research in 2013 academic year:

International workshop

1)    T.Endrjukaite, Y.Kiyoki. Music Data Retrieval Technique based on Instantaneous Frequency Spectrum. In the 5th International Workshop with Mentors on Databases, Web and Information Management for Young Researchers, Sapporo, July 2013.

 

International Conferences

1)    T.Endrjukaite, Y.Kiyoki. Music homogeneity analysis through instantaneous frequencies. In Proc of the 11th International Conference on Advances in Mobile Computing & Multimedia (MoMM2013), ISBN 978-1-4503-2106-8, Vienna, December 2013.

 

2)    T.Endrjukaite, Y.Kiyoki. Music Similarity Analysis through Repetitions and Instantaneous Frequency Spectrum. In Proc. of Conference on Signal Processing Systems IJSPS, ISSN:† 2315-4535, Sydney, December 2013.

 

3)    T.Endrjukaite, Y.Kiyoki. Musical Tunes Emotions Identification System by means of Intrinsic Musical Characteristics. International Conference on Information Modelling and Knowledge Bases; European Japanese Conference 2014, under submission.

 

International Journal

1)    T.Endrjukaite, Y.Kiyoki. Musical tunes continuous homogeneity analysis through signal decomposition and intrinsic modes frequencies processing. Journal of Mobile Multimedia, 2014, in process.