Aesthetic
Analyze of Computer Music
Teacher Maria Goga, Msc
School of Music Nr. 5,Bucharest, Romania,
email: goga_man@computermail.net
Postdoc Nicolae Goga
Eindhoven University of Technology, P.O. Box 513, NL-5600 MB Eindhoven,
email: n.goga@tue.nl
Abstract
The improvement of the aesthetic of the melody line of the music generated by the computer is a big challenge for the scientific world today. The research presented in this paper is trying to give a solution to this problem. A first experiment consisted on the following: 1) a tool for music generation was created; 2) some simple melodies for Bass were created and all the possible solutions for Tenor, Alto and Soprano were generated;a musical analyze from an aesthetic and musical form points of view was made; 3) a set of rules were derived from the analyze such that the aesthetic of the melody line of the music generated by the tool would improve.Out of this first experiments we derived a theory for feelings based music generation. Examples generated according to this theory will be also presented in this paper .
1. Introduction
The improvement of the aesthetic of the melody line of the music generated by the computer is a big challenge for the scientific world today. There are programs and mathematic theory [1] for computer music generation. Usually, this music is produced in conformance with the laws of harmony [3]. But when a computer scientist gives to a musician some music originated from the computer too often the musician is saying: I do not want only a music which is harmonic good, but also a music which will sound good. On another hand, for a mathematician, only employing the harmonic laws for the computer music generation still leaves a lot of freedom for the notes generation, and, consequently the number of possible solutions (for a simple leading melody) is usually very big. For example, when Bass is given, and its melody has a small size (13 notes), the computer can still produce millions of solutions for the melodies of Tenor, Alto and Soprano (all the solutions being harmonic good).
The research presented in this paper is trying to give a solution to the problems above mentioned. The team was formed by researchers from music and computer science. In main the experiment consisted on the following: 1) a tool for music generation was created; the tool implements the laws of classical harmony; 2) some simple melodies for Bass were created and all the possible solutions for Tenor, Alto and Soprano were generated; a musical analyze from an aesthetic [2] and musical form points of view was made; the principles of the classical music for a choir were engaged in the analyze; 3) a set of rules were derived from the analyze so that the aesthetic of the melody line of the music generated by the tool would improve.
This was a first experiment concerned with the computer music aesthetic. In our later experiences we took into account the set of rules derived from our first experiences. From this experiences we derived also a theory for feelings based music generation. The concepts of this theory are also applied and presented in this paper.
The paper is organized as follow. Section 2 briefly presents the harmonic laws implemented by the tool. Section 3 describes the tool. Section 4 presents the first experiment related with the aesthetic music analyze. Section 5 deals with the theory of feelings based music generation. Section 6 gives examples of applications of this theory. These examples are analyzed also esthetically. Section 7 outlines the conclusions.
2. The harmonic law
The music is a revelation higher than any wisdom and philosophy (Ludwig
van Beethoven)
The music (the sound art) revealed in the time of the centuries to be a main component of the human culture. The musical art has a theoretical system which is composed by a number of disciplines with different visions. In this way the musical theory, the harmony, the polyphony, the forms, the instrumentation and orchestration, the musical aesthetic and the composition, treat step by step the aspects studied by this art. From the multitude of disciplines enumerated above, the present paper is willing to `touch' only some aspects of the classical harmony, aspects which today can be used by the computer.
The chord -- the principal element of the harmony The chord is a sonorous unit formed by more simultaneous sounds [5] with a sonorous structure and aesthetic structure well organized.
In the natural resonance of the sounds the fundamental is repeated the most, the third is on the second place and the fifth is on the third place. It is not an accident that the sounds do (the fundamental), mi (the third) and sol (the fifth) represent the constitutive elements of the major chords.
The chords and the harmony are linked to the occurrence of the tonal system.From the necessity that the musical sounds must support each other, must have a weighting center, the idea of tonality was born. From the sounds of a tonality, a chord can be built on every musical degree. A chord is formed from third overlaps. Every sound takes the function given by its correspondent degree.
From the point of view of their functions, two kinds of chords can be distinguished: principal and secondary. The degrees of the principal chords are: the tonic (I), the subdominant (IV) and the dominant (V). The rest of the degrees are secondary : the supertonic (II) , the mediant (III), the submediant (VI), the leading tone (VII).
The harmony for four voices The present work treats the human voice harmony, especially the harmony of the principal voices -- Soprano, Alto, Tenor and Bas. According to their setting in the choir assembly the voices are divided in external voices (Soprano, Bas) and internal voices (Alto, Tenor). Usually, the melody is given by Soprano. Sometimes the other voices contribute also to the melody. The most used chords are the ones with 3 sounds (triads) which are distributed to 3 voices. The fourth voice doubles the tonic or the fifth (the third is doubled rarely).
In some chords some elements can be omitted. The tonic can not be omitted because it is the one which generates the chord. The same holds for the third because it gives the scale: minor or major. Consequently, the only element which can be omitted is the fifth.
The distance between voices For ensuring a homogeneous and balanced sonority a maximal distance should be respected among neighbour voices. For example, among Soprano and Alto the maximum distance is of an octave.
The melodic movements of the voices For having a cantabile song, some intervals, difficult to be singed, are not allowed:
The harmonic movement of the voices In the melodic movement we observed the horizontal development of one voice, but in the harmonic one we will observe two voices which moves simultaneously.The following movements between voices exists (self-explanatory by name): the direct movement, the parallel movement, the oblique movement, the counter movement and the counter-parallel movement.
The direct fifth and octave For the external voices, the direct fifth and octave are allowed if and only if the superior voice goes gradually.
The parallel unison, fifth and octave It is not allowed to have two voices which go in parallel in unison, fifth and octave when the voices go through two different chords.
The counter-parallel fifth and octave The succession of fifth and octaves between the same voices are not allowed for a counter-parallel movement.
3. A tool foor music generation
In this section we will present the tool for music generation. The tool which is especially built for our experiments can be used in the following way: a Bass melody is given as the input for the tool and the melodies of Tenor, Alto and Soprano are generated such that the constraints imposed by the harmonic laws are respected. However, the tool can easily be updated so that it will help the user to compose melodies for all the voices (Bass, Tenor, Alto and Soprano ), melodies harmonically good. The tool can be easily be adapted to be an interactively process in which the user gives notes for Bass, Tenor, Alto and Soprano to the tool and the tool tells to the user whether his/her choice is correct or not harmonically.
The main engine of the tool is the procedure Verify. This procedure has as parameters four arrays of notes (for Bass, Tenor, Alto and Soprano). The procedure verifies whether the musical composition formed by the notes of Bass, Tenor, Alto, Soprano is harmonically good. First it checks whether the distances between the notes of Bass and Tenor, Tenor and Alto, Alto and Soprano are in the allowed intervals (see Section 1). The next steps is to apply the procedure VerifyVoices to all the possible combinations between the four arrays of notes for Bas, Tenor, Alto and Soprano. The first 3 steps of the procedure VerifyVoices is to verify that there is not a direct octave, fifth or unison between two voices. After that it checks that there is not a parallel octave, fifth or unison between two voices, at the end it looks for counter-parallel octaves or fifths.
As explained above the constraints imposed by the harmonic law are checked out by the procedure Verify. The generation of the notes Tenor, Alto and Soprano is done by a back tracking procedure. The algorithm generates notes for Tenor, Alto and Soprano in a size equalling the length of the Bass melody. First, it fills the first position by generating three notes for Tenor, Alto and Soprano. After that it goes to the next position and generates again notes for Tenor, Alto and Soprano. This is done till a complete melodic solution is constructed. At each position the back-tracking algorithm verifies whether the melody generated is harmonic good (this is done by using Verify). After a first solution, the algorithm generates other solutions. It stops only when all the possible combinations for every position are tried and, consequently, all the possible solutions for the melodies of Tenor, Alto and Soprano are produced.
The tool have also the possibility to accept pattern functions for the Soprano voice. An example of such a pattern is: the voice is moving gradually. A first use of the patterns is to generate music when the pattern and the Bass are given. Another possibility is to ask the tool to generate all the voices when a pattern of the Soprano is given. This option was used for the feelings based music generation experiment.
4. The first experiments
The melody is the most expressive way of communication of the music. The melody is a general reference for the compositional, aesthetic, theoretical and historical characteristics of the sound art. The melody is the center to which all the characteristics of the musical image converge.
A melody can be well composed harmonically but, sometimes, it does not sound well. In the work presented here, there are many musical examples which confirm this truth. The same Bas can generate, on the basis of the harmonic laws, melodies which are correct from an harmonic point of view but which do not sound well. More precise, these melodies do not have a beginning, a culmination and an end.
The Bas given (imposed) is in a pedagogical style, arranged in such a way to make easy the analyze of the melodies generated. The harmonic laws of the classical style are used. From this reason the analyze of the melodies are made also from a classical point of view. The Bas of each of the three melodies which are used in the first experiments is formed from the principal degrees of the do major scale. The rhythm is the element which differentiates the melodies.
Because the number of solutions generated was too big we considered randomly the same number of solutions for each melody, we made observations related with the aesthetic problems encountered. What is interesting for us is to observe whether the melodies created for Soprano (and for Alto and Tenor) can be interpreted and whether the aesthetical melodical curve which keeps the idea of a beginning, a culmination and an end exists. We call correct melodies (37% in our case) the ones which correspond to the harmonic laws and the aesthetic laws.
There are exceptions from the aesthetic rules, melodies which do not respect strictly the `melodic-aesthetic` curve which gives the criterion of correctness. In a number of cases, the culmination is at the beginning, or at the beginning and at the middle, or at the middle and at the end, or at the end (46% exceptions). It is common to have the culmination only once (around the middle of the melody ), for such a small size.
Another problem encountered is that some melodies from the exception group can not be easy interpreted, are not easy-cantabile because the culmination (the notes in the acute register ), bypasses the admitted limit (goes too high), and, consequently, the other two support voices, Tenor and Alto, go also too much in the acute.
So called non-correct melodies can not be called melodies, because they consist of one or two sounds which are repeated continually and we can not speak about a beginning (in the medium register of the voices), a culmination (in acute) and an end (in medium).
5. Feelings based music generation theory
In the previous chapters we presented a tool, aesthetical and classical musical form analyses and general rules of improvement (derived from the analyses performed) for the computer music. In the current chapter we propose a new theory for the music generated by the computer. We experimented with it on the tool presented in this paper.
When a composer is involved in his compositional process, especially when he/she tries to compose music for different texts, the composer wants the music to awake different feelings in the hearers of the music. The feeling can be decided by the composer or can be derived from the text. For example, a piece of text expresses joy, another one violence or peace, etc. On another hand it is a known fact that using some general rules for composing music can produce in the audience some reactions. For example, repeating a small sequence of notes many times make a hypnosis state to be induced. The joy can be produced in many ways; for example by a melody which goes upwardly and gradually.
We propose to generate music according to a similar process used by the composers. Our method applies especially for texts which need to be sustained by melodies. When looking to the text, the human (a composer or a non-specialist) decides each portion of the text which feeling needs to awake. The feeling can be derived from that portion of text or can be decided by the human. The computer receives as an input data the set of feelings and for each feeling the length of the melody which needs to be generated. Now, from its data base, for each feeling from the set, the tool will choose rules or a pattern which express that feeling and will generate music, till the whole melody is constructed. The human will receive back the set of melodies generated and he/she can choose one of these or get inspiration from this set in his/her compositional process.
Of course a key element of the whole process is the knowledge (the data base) which the tool have related to the melodically patterns which induce a given feeling. In the current paper we worked out a list (which is not exhaustive) containing feelings and their associated pattern. One important observation is that one feeling can be awaked by more patterns. The tool has the possibility to choose from more patterns for a given feeling, therefore its creativity is enlarged.
The list of sentiments and their correspondent
patterns are given below. This list is not exhaustive and it was worked out by
the authors of this article with the help of the existing literature (see for
example [5]).
1. Restlessness:
a) Gradually descendent movement combined with ascending movement
having short breaks which suggests the restlessness idea.
b) At the
beginning of the melodic sequence gradually movement followed by jumps of
fifths or sixths; the melody is formed by the repetition of this melodic
sequence.
2. Peace
a)
Gradually
ascending movement combined with a
descendent jump of a fifth
b)
Gradually
ascending movement followed by a descendent jump of a third combined with a ascending
jump of a sixth and after that again a descending jump of a third
3. Consolation
a)
Gradually
ascending movement combined with gradually descending movement
4. Innocence
a)
Gradually ascending movement which contains a short musical proposition which will be combined with the same
musical proposition transposed to a range of a fourth.
a)
Gradually
ascending movement combined with a gradually
descending movement and everything repeated (short values for the times
of the notes)
5. Delicately
a)
Jump
of a sixth at the beginning followed by a gradually descending movement
combined with a gradually ascending movement (short values for the times
of the notes)
b)
At
the beginning the repetition of the sound once followed by a jump of a sixth
combined with descending jumps of thirds and ascending movement in jumps (short
values for the times of the notes)
6.
Sadness
a)
The
first sound repeated five or six times combined with gradually ascending
movement
b)
Gradually
descending movement followed by descending movement in jumps combined with
ascending movement in jumps (large values for the times of the notes)
7. Trust
a)
Ascending
jump of a fourth followed by a repetition of the same sound (three or four
times) combined with the repetition of a sound obtained from a ascending and
gradually movement (middle values for
the times of the notes)
b)
The
first sound repeated twice or three times combined with an ascending and
descendent movement with jumps of
thirds (middle values for the times of the notes)
8. Love
a)
Gradually
ascending movement followed by gradually
descendent movement combined with jumps of fifths and followed by the repetition of the same sound
b)
Descendent
movement in jumps of thirds combined with ascending movement in jumps of thirds
followed by large values of times on the same sound
9. Joviality
a)
At
the beginning jump of a sixth followed by gradually ascending movement combined
with an gradually movement or movement with jumps of thirds
b)
Gradually
ascending movement followed by descendent
jumps of sixths or sevenths
combined with jumps of fourths followed by a gradually descending
movement
10. Joy
a)
Descendent
movement with jumps of thirds (unison for all the voices) combined with
ascending jumps of thirds.
The music is a strange and subtle form of art.
It is formed from different
ingredients: rhythm, harmony, melody, their mixture, etc. In function of
the presence and the equilibrium of each of these ingredients the music can be
peaceful or invigorating, profane or noble, philosophical or orgiastic. The
music aims to influence the systems of our emotions.
It is interesting to observe and to analyse
from a medical point of view how the musical patterns listed above induce their
correspondent feeling (listed also above). But the way in which the musical
sound from its entrance in the external hear induces the feeling production is
not in the focus of the current article. Therefore we will explain this only
for the variant a) of point 1) above: how the musical pattern induce the
feeling of restlessness.
This musical pattern has short breaks at the
beginning of the three cumulated values which go always from the acute register
to the grave register. Another characteristic is the syncope which is used
repetitively. The syncope reverses the natural rhythm of the body which induces
a sensation of anxiety which conduct to restlessness.
According to the medical experts [4], this musical pattern
modifies the respiration and the blood circulation. This can lead also to the
modification of the percentage of the sugar in blood. Under the influence of
this music the pupil reflexes and the brain circumference can be changed. These
modifications provoke anxiety, agitation, leading to restlessness.
We worked out examples according to this theory with the tool described in the previous chapters. In the examples worked out, we choose some texts, we decided the feelings and we gave them to the tool together with the associated lengths. The tool produced music, which was analyzed from an aesthetical and musical form point of view. These are presented in the next chapter.
6. Implementing the
feelings based music generation theory
In the current
article we showed musical patterns based on ten types of feelings. For every
feeling we presented two musical variants which induce that feeling. For the
feeling of happiness we worked out an example. This is based on a text from the
book Psalms of the Bible (Ps. 148, 1)
This text expresses feelings of praise for the
Lord, happiness from the deepness of the heart. These feelings can be induced
by different patterns. A first pattern is one in which the melody of Soprano
has a jump of a fourth in the beginning followed by a gradually and descendent
movement. The second pattern for Soprano has the culmination at the beginning
of the melody (in acute) followed by a
movement which has jumps of maximum of thirds.
The
jump of a fourth for the first variant expresses the joy found in the text. The
gradually descendent movement suggests the trust in the Almighty God which
embraces gradually the soul of the composer (listener). The culmination of the
second variant expresses an explosion of joy. This is followed by a gradually
stabilisation of the trust in God (expressed by the movement which has small
jumps).
Praise the
Lord from the hea vens
We succeeded to generate melodies for both
patterns. Both melodies has a nice sonority. The harmonic law implemented by
the tool are respected. We did not encounter aesthetic problems.
7. Conclusions
We created a tool for music generation. The tool is based on a backtracking algorithm which explores all the possible solutions and chooses the only ones which respects the constraints
imposed by the harmonic laws. When the melody of Bass is given it can generate all the solutions for the melody of Tenor, Alto and Soprano. This functionality of the tool was used for the aesthetic music analyze of our first experiences. The tool can be used also for imposing a pattern to the Soprano voice and to generate all four voices for a given pattern (or sets of patterns).
In our first experiments we used melodies of small size (8 -- 13 notes) which were made in a pedagogical style, with imposed bass. For this experiment, we used chains of major principal chords. Many solutions given by the computer were distributed in the acute register which for a professional choir is not always easy to be intonated; for an amateur choir this is almost impossible. We also observe from the examples generated that the melody does not have a musical form with introductive part, culmination and ending of the musical phrase. So we found the culmination in the beginning and sometimes remaining in the melody movement. In many examples, the soprano voice remained on the same notes which does not correspond from the aesthetic or musical form points of view. These are only some negatives aspects met in the examples produced by the computer.
Out of these experiments we derived a theory feelings based music generation. We employed the principles of this theory for generating music and we analyzed it from an aesthetic point of view.
We will continue the study of the quality of the music generation in other domains (such as for a specific instrument). We will implement also other kind of harmonic laws such as the modern and contemporary harmonies and improve the quality of the computer music. Another possible fellow-up of this research is to observe how such tools can be used efficiently in the domain of education, by teachers and students.
References
[1] Jackendoff, R. S. and F. Lerdahl (1996). A generative theory of tonal music. Publisher: MIT Press.
[2] Leopoldo, H. (1971). Introduction a la estetica de la musica. Publisher: Editorial Paldos, Argentina.
[3] Schoenberg, A. (1900). The theory of music. Germany.
[5] Schullian, D. and Schoen, M. Music and Medicine. USA.
[5] Wyatt, K. and Schroeder, C (1998). Harmony and Theory. Publisher Hal Leonard, USA.