Hofmann-Engl (2001, 2002a, 2002b, 2003a, 2003b, 2004 & 2005) introduced melodic similarity as a combination of melotonic (pitch), chronotonic (durational) and dynamic (loudness distribution)similarity (compare the research resource page).

In 2005, Robert D. Vincent at MC Gill University, Toronto, published a similarity tutorial coining the term Hofmann-Engl similarity referring to melotonic similarity. The tutorial's home page is here.

The computations of the applet in Toronto however, appears to contain some errors and does not provide the correct answers.

Hence, below an applet which does compute the Hofmann-Engl similarity correctly. However, the user interface is not very friendly at this point in time.

Firstly, not melota/pitches are to be entered, but exact frequencies. Still, there are a number of good frequency-pitch tables available online and particularly this one.

Secondly, the values for k₁ and k₂ are unclear to a new user. The preset values are the best values as found within one experiment by Hofmann-Engl (2003). It might be an idea to simply keep the preset values unless the user is familiar with the theory behind the algorithm.

Thirdly, there is no proper error protection. For instance, if a user forgets to clear a previous calculation, the following one will be wrong.

Finally, the fragments have to have the same length. For instance, a melody of 5 notes can be compared only to another one with 5 notes. Comparing 4 notes with 5 will result in wrong computations.

Nevertheless, this applet at least works even if there is much room for improvement. Enter the frequencies for fragment 1 (the text will change from 1.1 to 1.2 to 1.3 etc) and then do the same for the fragment 2. Once both fragments (melodies or chains) have been entered the similarity can be computated.

The program is copyright protected. However, the source code can be found here. The author welcomes suggestions.

---

Chameleon Group of Composers © 2020