Artificial Intelligence – music analysis example using Guitar

Introduction

Human Resources has seen a huge jump as a profession, taking up leadership role, more than ever before. In this article we will indulge in a quick peek of some music analysis example, to point out that Artificial Intelligence in HR is necessary. We further develop a Skill Evaluator using Fuzzy Logic.

HR as seen by a lot of people these days is about:

• Hiring and Recruitment.
• Salary and Compensation.
• Reviews and Performance Appraisals(For ranking employees).
• Training and Development.

Given that these are just the tip of the iceberg with respect to the activities that HR is supposed to conduct, but at this stage it is easy to ask AI to do everything, and that would be akin to artificial intelligence for dummies and we would be taking on more than what we could chew. Therefore, it becomes easier to use the above mentioned items as broad based yardsticks for conducting HR business using AI. Recruitment implies to hire fresh workforce or for existing employees to move up a role, with cross hiring, as also, whenever we hire we must allocate accurate salaries and compensations, at the year end or even bi-annually or quarterly, Reviews and Performance appraisals must be conducted to map performance, this is perhaps the most useful when employed with the skill evaluator (what is it? More on this later in this article) using music analysis example.  Finally, for employees to perform, they must be trained systematically.

All of these 4 bullets above are complex activities, and these days with the coming of Artificial Intelligence, it is common to use AI to solve quickly for these HR activities, and as mentioned earlier we should avoid falling into the trap of artificial intelligence for dummies, where AI is used like an HR mantra.

An HR quick peek would reveal,  AI software to be effective requires a real life skill to evaluate the 4 points mentioned above.  When we say real life or real world, we imply something that is “Not of software,” but something that can be taken as an input by AI software, something like music played on the guitar.  It would be teaching Artificial Intelligence to dummies, unless a musical analysis example was given.

Basics of Music scales

This section requires a rudimentary understanding of music theory, skip to the next section if you are a newbie.
Guitar Music scales have an enormous potential to be the inputs to an Artificial Intelligence software, a G-major, for instance, is made up of the following 7 notes:

G. A. B. C. D. E. F#

where # implies sharp.

A music analysis, which means to look closely into our music analysis example (we will get a sample audio later in this article), of the scale being played on the guitar will generate different notes as compared to when a C major– C D E F G A B, A major – A B C# D E F# G# or D major – D E F# G A B C# etc is played.

Also within the same scale, say G-major, different employees whose skills are being evaluated with skill evaluator using AI, will fare differently as will be evident by the quality of notes that the guitar will output for their work.  We could use this to evaluate all the above 4 core points of HR business (Hiring, Salaries, Performance appraisals and Training). An A+ player or employee will have better quality of music as compared to an A, B+ or B player.  This would hold true for all the aspects of our music analysis example.

Guitar generated music

To test whether we could do this, we planned the following guitar solo piece with two people who are of the category of junior management and CEO level. The piece is as played on the guitar, which I will endeavour to record in it’s most perfect form, as well as what the two people worked up when they worked and we played the guitar to their work.  The fretboard for all notes on the guitar and for G major scale is as displayed below in figure 1.1 and figure1.2:

Fig 1.1 : Complete Notes on the fretboard

Fig 1.2 : G major Notes on the fretboard (diatonic scale)

Without going to much into the theory of music, suffice it to these following points:

1. The music analysis example is played in G major scale.
2. The notes of G major are as seen in fig. 1.2.
3. Any change in notes will be obvious in the recording.
4. The piece is played on the 4th fret to the 7th fret of the guitar. (each vertical separation on the guitar is called a fret)

The test of Skills

The scores were calculated on the following skills for junior management:

1. Cohesiveness.             :    Verbal ability (VA)
2. Comprehension.           :    Comprehension ability (CA)
3. Fluency.                        :    Reasoning ability (RA)
4. Clean strings.                :    Aptitude

And for CEO level they were calculated as follows:

1. Cohesiveness.      .       :     Finance and P&L
2. Comprehension             :    Mergers and acquisitions
3. Fluency                         :     Leadership
4. Clean strings                 :     Core business goals
5. Note awareness            :     Mentoring
6. Composition                 :     Vision/Mission

Each candidate was tried out thrice and the best score taken as their actual score.  Candidate 1 is CEO level and candidate 2 is junior management level.

Below appear the audio files of the recording for music analysis example.

Candidate 1

Candidate 2

 

As you hear this music analysis example and, this is an analysis of music by ear, you can note the difference in the two works.  Music when analysed this way is left to the fine tuning of the ear of the listener, in future articles we will explore using Artificial intelligence for analysing this and the parameters used for AI.  Both candidates were doing their work when we did a music analysis example recording, meaning this is live data.

 

Why Guitar as the instrument for music analysis?

I have been experimenting with AI music analysis examples, that is reading the imprint, the code of it, of guitar generated music for HR practices, from a practical angle it has enormous potential.  This is primarily because of the following points:

• Accuracy and precision.
  Accuracy is the percentage times we hit the skill measurement target.
• Crossover between Music scale and HR activities.
  Co-relation between Music scale and HR activities is easy to draw.
• Logical progression of Music scale.
  There is a lot of logic in music and this can be applied to HR.
• Complex music patterns to handle all the above items.
  Music provides and possesses enough complexity to handle even complex tasks like Performance appraisals.

What we are saying is that music played on the guitar if incorporated with an AI software will be useful for accurately measuring skills.  This is the composition of the Skill Evaluator.

So then the first step is to find the parameters we must evaluate Skills on, and these may be classified as below:

• Hiring Freshers or Junior management personnel
  • Verbal ability.
  • Reasoning ability
  • Aptitude.Instead of recruiting and placing based on the. CV and whatever little experience they might have at this level we hire for the above, they must have skills to learn new skills that would be required at work.
• Hiring Middle management personnel.
  • On-the-job skills required.
    • Dissecting content of documents.
    • Ability to document.
    • Plan and execute.
    • Delegation.
    • Communication.
    • Presentations
  • Subject matter knowledge
    • Understanding of the core business.
    • Understanding finance.
    • Understanding of supervision.
    • Ability to generate profits
  • VA / RA / A as above
• Hiring Senior management
  • On-the-job skills
    • Mergers and acquisitions.
    • Finance and profit & loss implications.
    • Vision and mission.
    • Mentoring.
    • Leadership
  • Subject matter knowledge.
    • Core Business goals.
    • Achieving targets.
    • Salaries and Bonuses

With these points in focus, Skill Evaluation becomes a question of benchmarking guitar playing skills against the above mentioned parameters for employees, as you would note if we were to evaluate skills instead of experience, and though experience is important (and we will evaluate it too in other posts), we are better placed at getting an employer – employee fit.

How to check Music

(using the music analysis example)

Developed in the 1960’s and said to have enormous potential, Fuzzy Logic never picked up beyond some washing machines and TV applications, it just wasn’t there when it came to artificial intelligence and machine learning.  Though, I must admit, I think the world of it, in my last application of it, I had used wood working to generate satisfaction level with the employee’s work (annually, bi-annually or quarterly), it was easy to understand and work with.

This time around the scope Is bigger and the project much tougher with the Guitar, music analysis has been my pet project for about two months, and it is only now that I am taking on the onerous task of writing code for the project.  Music analysis requires the analysis to be along the below points:

1. Melody

• Frequency & Pitch                            :   Changes in pitch of sound.
• Dynamics, Intensity and Loudness    :   Dealing with the loudness of sound.
• Timbre                                             :   The tone or colour of the music or sound.
• Structure                                          :   The structure of music being played if there is one.
• Scale                                               :  Whether all notes in the scale have the right freq.

2. Harmony

• Euphoric : G major is happy and lively
• Sad : Sad as is obvious
• Happy : Happy tone
• Tense : Tense feel to the music

Process of Music Analysis

It seems to be tough task at the onset, unless you begin to split the project into Melody and Harmony sections.  As a first, I will look at Melody.

The first 4 items of Melody are connected to each other, and the fifth item, Scale, and it can be found out for any sample audio file too, and so also the notes comprising of it using python Librosa library. What we want to do effectively is take the original or the most perfect piece and check out its Frequency & Pitch, Dynamics, Intensity & Loudness, Timbre, Structure and Scale.  Then knowing this, compare with the pieces that were a result of the candidate/employee working as per our music analysis example.  The compare operator will give us our scores on a scale of 5.  Do the same for all the other five items as above and we would have a cumulative score for the candidate, the more closely these scores match the perfect piece score, the more likely we are to hire them, increase their salaries and give promotions etc.

Music Analysis is a software heavy task, though it is seen sometimes as what humans do mentally when listening to music, and that too easily. The software that has to be developed by us incorporates Python Librosa library, explicitly meant for music analysis.  Even the scale analysis which is the 5th point in Melody can be analysed this way.

Conclusion

The python Librosa library is the library to code with for all that we intend to do.  Once the pitch has been found out, that is the frequency, a comparison will be drawn between the perfect piece and candidate pieces for contour, range, interval, structure and scale.  The comparison to be drawn will use Fuzzy Logic (more on this in later articles), to derive the rating on a scale of 0-5.  Also each employee being evaluated will need to be given at least 10 minutes of guitar time for an accurate and precise rating.

References

 1,268 total views,  1 views today