README.md

In lizhongliu1996/melodyExtracter: This is a package that can remove the vocal part of a song that sings the melody, gives you only the accompaniment so that you can record and sing on karaoke or live

How to install

run devtools::install_github("lizhongliu1996/melodyExtracter") in R console

melodyExtracter

Having trouble find the accompaniment of a new song which you prepare for singing to your classmates or colleagues?

Tired of transcribing a song by repeat listenning to it thousands of times?

Sick of spending so much time on clicking notes on digital work station and finding the matched tone color of each instruments?

Based on some acoustic technologies, it is possible that we can save you much efforts through our algorithm.

Usually a typical commercial song is composed of many tracks including vocal, piano, percussion, guitar, orchestra and sometimes electronic synthesizer. According to the definition of "entropy", a parameter to reflect the degree of chaos, entropy is constantly increasing, and the increasing process is irreversible. For example, it is easy to cook a dish with recipe, but it is pretty hard to know the ingredients and spices given a cooked dish without recipe. Based on that, after rendering many tracks into one single wav or mp3 file, it is a pretty hard challenge to remove a track, so how to get ride of the melody is quite a problem especially when you cannot find the midi file online.

Luckily, French mathematician Fourier demonstrated the rule that any complex waveform can be seen as a superposition of several sine waves. The Fourier transfrom is a powerful weapon that can easily separating the sine waves which make up the composite waveform and transform them into a spectrogram, which we can easily see the characteristics of composite waveform and do some manipulations to get the desire waveform. Sound is one type of mechanical wave with salient features, thus we can use Fourier transformation to play some interesting tricks in music.

This package refers to the principle of how the headphone brand "BOSE" remove the noise by adding the polarize reverse wave of specifical frenquency we want to get rid of, however rightnow we can only deal with song that has mono vocal part with 0 pan.

lizhongliu1996/melodyExtracter documentation built on May 16, 2019, 4:09 a.m.