It is well known that music can convey emotion and modulate mood, to retrieval music by mood is sometimes the exclusive manner people select music to enjoy. This paper concentrates on music retrieval by detecting mood. Mood detection is implemented on the viewpoint of Computational Media Aesthetics, that is, by analyzing two music dimensions, tempo and articulation, in the procedure of making music, we derive four categories of mood, happiness, anger, sadness and fear. Concretely, with regard to music in the format of raw audio, after tempo is detected using a multiple-agent approach, a feature called relative tempo is calculated, and after the mean and standard deviation of the feature called average silence ratio in the presented computational articulation model are calculated, a simple BP neural network classifier is trained to detect mood. Users retrieval music by browsing the 3D visualization of feature space associated with specific mood. This paper reports the experimental result on a test corpus of 353 pieces of popular music with various genres. 1.

Abstract not found

Figure 1. Stills from "“Concerto for Virtual Strings and Faces” " animation created by remapping affective data from a music score. We propose a method to extract the emotional data from a piece of music and then use that data via a remapping algorithm to automatically animate an emotional 3D face sequence. The method is based on studies of the emotional aspect of music and our parametric-based behavioral head model for face animation. We address the issue of affective communication remapping in general, i.e. translation of affective content (eg. emotions, and mood) from one communication form to another. We report on the results of our MusicFace system, which use these techniques to automatically create emotional facial animations from multiinstrument polyphonic music scores in MIDI format and a remapping rule set.

Content-based prediction of musical emotions and moods has a large number of exciting applications in Music Information Retrieval. However, what should be predicted, and precisely how, remain a challenge in the field. We provide an empirical comparison of two common paradigms of emotion representation in music, opposing a multidimensional space to a set of basic emotions. New groundtruth data consisting of film soundtracks was used to assess the compatibility of these models. The findings suggest that the two are highly compatible and a quantitative mapping between the two is provided. Next we propose a model predicting perceived emotions based on a set of features extracted from the audio. The feature selection and transformation is given special emphasis and three separate data reduction techniques are compared (stepwise regression, principal component analysis, and partial least squares regression). Best linear models consisting of 2-5 predictors from the data reduction process were able to account for between 58 and 85 % of the variance. In general, partial least squares models performed the best and the data transformation has a significant role in building linear models. 1.

Abstract. Apart from genre- and artist-based organization, emotions are one of the most frequently used characteristics to describe and thus potentially organize music. Emotional descriptors may serve as additional labels to access and interact with music libraries. This paper reports on a user study evaluating a range of emotional descriptors from the PANAS-X schedule for their usefulness to describe pieces of music. It further investigates their potential as labels for SOM-based maps for music collections, analyzing the differences for labels agreed upon by a larger group of people versus strictly personalized labellings of maps due to different interpretations by individual users. 1

Abstract – This paper addresses the issue of affective communication remapping, i.e. translation of affective content from one communication form to another. We propose a method to extract the affective data from a piece of music and then use that to animate a face. The method is based on studies of emotional aspect of music and our behavioural head model for face animation. 1.

Introduction Music is not only for entertainment and for pleasure, but has been used for a wide range of purposes due to its social and physiological effects. Traditionally musical information has been retrieved and/or classified based on standard reference information, such as the name of the composer and the title of the work etc. These basic pieces information will remain essential, but information retrieval based on these are far from satisfactory. Huron points out that since the preeminent functions of music are social and psychological, the most useful characterization would be based on four types of information: the style, emotion, genre, and similarity [Huron,2000]. The relation between musical sounds and their influence on the listener's emotion has been well studied. The celebrated paper of Hevner [Hevner,1936] studied this relation through experiments in which the listeners are asked to write adjectives that came to their minds as the most descriptive of the music played.

Background in music theory and analysis. The energy potential of music unfolding in time is conditioned by its structural situation, conceived in the sense of the chief reagent of the realization of musical thinking and affective states. The special task of formal analysis is to discover the expressive musical feature (Kivy, 2002) which, in the given context, influences other features, i.e. to find out which musical feature assumes the role of the sign of recognition of the structure at the given moment (Popović, 1998), thus becoming the pillar of structural organization, its “point of gravity ” (Lerdahl & Jackendoff, 1983) in the unfolding of musical events. Background in music psychology. Due to the structure of music unfolding in time, expressive communication of a specific musical piece causes differentiated affective responses. But the question is: where is the key point of the experienced musical expression? Is the emotional process unfolding in the course of listening a function of the pillars of musical structure, or is it dependent upon the specific personality characteristics and experience of the listener? Aims. We aim to clarify the segmentation and the structural points of gravity of Mozart’s Phantasie, the emotional and narrative response pattern in relation to the segments, and their

There are many problems in the field of music information retrieval that are not only difficult for machines to solve, but that do not have well-defined answers. In labeling and detecting emotions within music, this lack of specificity makes it difficult to train systems that rely on quantified labels for supervised machine learning. The collection of such “ground truth ” data for these subjectively perceived features necessarily requires human subjects. Traditional methods of data collection, such as the hiring of subjects, can be flawed, since labeling tasks are time-consuming, tedious, and expensive. Recently, there have been many initiatives to use customized online games to harness so-called “Human Computation ” for the collection of label data, and several such games have been proposed to collect labels spanning an excerpt of music. We present a new game, MoodSwings

In this paper, the automated detection of emotion in music is modeled as a multilabel classification task, where a piece of music may belong to more than one class. Four algorithms are evaluated and compared in this task. Furthermore, the predictive power of several audio features is evaluated using a new multilabel feature selection method. Experiments are conducted on a set of 593 songs with 6 clusters of music emotions based on the Tellegen-Watson-Clark model. Results provide interesting insights into the quality of the discussed algorithms and features. 1