Quantcast

AI Algorithm Correctly Detects Subtle Changes In The Music Of The Beatles

July 27, 2014

redOrbit Staff & Wire Reports – Your Universe Online

Music enthusiasts have long recognized the dramatic changes the music of The Beatles went through in only a few years’ time, but now for the first time researchers have scientifically measured the legendary rock band’s progression using an artificial intelligence algorithm capable of analyzing and comparing musical styles.

Writing in the August edition of the journal Pattern Recognition Letters, a pair of computer scientists from Lawrence Technological University in Southfield, Michigan explained how their algorithm demonstrated that the structure of the Beatles music changed progressively from one album to the next. The study looked at 11 songs from each of the 13 studio albums released in the UK, and then quantified the similarities of each song and each set of songs.

First, the algorithm converted each song into a visual representation of its audio content known as a spectrogram, converting an audio analysis problem into an image analysis one. To solve the image analysis issue, the researchers applied comprehensive algorithms to convert each spectrogram into a collection of nearly 3,000 numeric descriptors that reflect visual aspects such as textures, shapes and the statistical distribution of the pixels.

Finally, pattern recognition and statistical methods were used to detect and quantify the similarities between different songs. The study authors, assistant professor Lior Shamir and graduate student Joe George, had previously developed and used audio analysis technology to study how whales communicate, and then expanded the algorithm to analyze the albums not only of the Beatles, but also of U2, Queen, Abba and other popular performers.

Following the analysis, the technique used by Shamir and George automatically placed the Beatles’ albums in their correct chronological order, starting with the first album, “Please, Please Me.” The computer algorithm determined that the songs on that album most closely resembled the second one, “With the Beatles,” and were least like the songs featured on the band’s final recorded album, “Abbey Road.”

The algorithm also placed early albums “Beatles for Sale” and “A Hard Day’s Night” along with the first two, then placed in order the albums “Help!,” and “Rubber Soul.” Those were followed by “Revolver,” “Sergeant Pepper’s Lonely Hearts Club Band,” “Magical Mystery Tour,” “Yellow Submarine,” and “The Beatles” (The White Album). Finally, even though “Let It Be” was the last album released by the band, the algorithm correctly identified that the songs it contained had actually been recorded before those featured on “Abbey Road.”

“People who are not Beatles fans normally can’t tell that ‘Help!’ was recorded before ‘Rubber Soul,’ but the algorithm can,” Shamir, corresponding author of the study, said in a statement. “This experiment demonstrates that artificial intelligence can identify the changes and progression in musical styles by ‘listening’ to popular music albums in a completely new way.”

He added that he believed that this type of research will be historically significant: “The baby boomers loved the music of the Beatles, I love the Beatles, and now my daughters and their friends love the Beatles. Their music will live on for a very long time. It is worthwhile to study what makes their music so distinctive, and computer science and big data can help.”

The algorithm was similarly successful with its other musical experiment, with one unusual exception, according to the researchers. As they explained, the technique found more similarities in two Tears for Fears albums released 15 years apart – “Seeds of Love” from 1989 and “Everybody Loves a Happy Ending” from 2004 – than those released by the group’s principal songwriter, Roland Orzabal, following their temporary disbandment in 1991.


Source: redOrbit Staff & Wire Reports - Your Universe Online



comments powered by Disqus