Find A Song Title Just by Humming The Lyrics

By Ryan Dube

Have you ever found yourself having one of those mind-numbing moments when you’ve fallen in love with a new song that played on the radio a few times, but you can’t remember the title or the lyrics?

Here at MakeUseOf, we’ve focused on a number of websites and tools that you can use to find a song by its lyrics, such as Brad’s article on the top 5 sites to find song lyrics online or David’s article on how to find song titles with just a few lyrics you do know. Today, I’d like to cover a very new solution that’s available for the most common scenerio – when you don’t know the title of the song or any of the lyrics, but you can hum or whistle the song just fine!

Voice Recognition Technology Saves The Day

There are a number of cutting edge computer technologies that have always fascinated me, mostly because they represent areas where computers are finally entering into the realm of artificial intelligence. Two of those include image recognition and voice recognition.

While I’ve not followed the state of the art of each field, I’ve always been under the impression that both areas of computer intelligence have a very long way to go. However, what would you say if I told you that I’ve discovered a computer program, an online application to be exact, that you could sing, hum or whistle a tune to and it could accurately guess the title of the song. Images of science fiction movies, where starship crew members seek the help of the ship’s AI computer system come to mind. Well, it’s no longer science fiction – this technology is now a reality at Modomi.

Midomi itself isn’t exactly a very special or unique website. Just like every other website where you can search for and purchase music online, Midomi lets you research music, bands and more. You can watch music videos, listen to music clips or even join the Midomi fan club for any particular music group. However, the true heart of Midomi becomes apparent (and the site absolutely comes alive) when you start becoming “fans” of particular groups. As you enter fan pages, you’ll become aware of a Facebook-like social network of music lovers. As I was browsing through this site, I noticed something rather unique compared to most other music websites. Midomi once again takes advantage of interactive audio by allowing fans to make and upload their own recordings of the music that they’re fans of.

As a fan, you can listen to any of these uploaded recordings. I joined the Green Day fan club and immediately saw user dude34bja’s recording of When I Come Around at the top as the most listened to recording. I clicked on it and was blown away by her fantastic acoustic rendition of the song. You too can record your own version of any of your favorite songs if you click on the “Studio” menu option at the top of the Midomi page.

Record your songs, and if your good enough you’ll very likely develop your own fan base right within the Midomi online community. Get enough votes and you actually appear as a “Midomi star.” In addition to recording music, the Midomi site is just completely integrated with audio at every turn. You can even upload audio comments at the site instead of typing them.

Now, all of this is very cool – but what about the whole reason we came to the site? That little box at the top of the first page that reads, “Click and Sing or Hum.” Does it really work? I decided to test the accuracy of this online application by humming Losing My Religion, by R.E.M. After clicking the text box, the recording began.

I hummed the first 16 seconds or so of part of the song – no words whatsoever – just straight humming. I was absolutely convinced there was no way that this application would be able to recognize the song without a single line of lyrics. When I was done I clicked on “Click to Stop” and waited for the results. This is what came up.

The top “guess” was a perfect match. Unbelievable. I figured it was a fluke – a completely lucky guess, so decided to try it again with a more difficult tune that wasn’t so modern – I went back and hummed the first 15 seconds of “Over the Rainbow,” sung by Judy Garland in the Wizard of Oz. I chose this song because there is absolutely no way I can hum it perfectly in tune. After 15 seconds of forcing the recorder to suffer through my cracking and screeching as I attempted to hum some of the tune, I clicked stop and grinned. No way it’s going to know what that horrid noise was. The following were the results.

I have to say, if anything convinced me – this did. As far as I can tell, you only need to know the general tune of a song – no title and no lyrics required. This amazing little Midomi web app will remind you of the title after just a few seconds of humming. Even better – you can download the app to your iPhone and carry around this little song-title-reminder app with you everywhere. Hard to believe? Visit the site and give it a shot – and let us know in the comments section below whether it blew you away. I know it did for me. Have fun!

Voice Translator Coming to iPhone and Blackberry [VID]

BY Cliff Kuang

Witness the Real-life (Beta) Version of Star Trek's Universal Translator

Sakhr Mobile develops an iPhone app for troops and diplomats that translates Arabic speech into English, and visa versa--and there's video of it, in action.

Fast Company just got a tip about one of the first products of that effort, an iPhone and Blackberry app, created by Sakhr Software and Dial Directions, that does natural language translation. Check out the video of the app, translating sinister-sounding--but all-too-real--phrases like "The prime minister will form a new government":

This is the first product from Sakhr, which has just announced its acquisition of Dial Directions. Sakhr works on devices that scan Arabic words and then offer machine translations, and its customers already include the U.S. Department of Defense, U.S. Department of Homeland Security, and the U.S. Department of Justice. Dial Directions, for its part, creates voice-entry applications.

Creating the app you see above was a matter of combining those technologies--first, the speech recognition part, which transcribes what's being said, and then applying Sakhr's know-how in translating the text that's produced. Zoinks! The future, apparently, is now. And while the military applications are obvious, we can't help but think the commercial applications are far more vast. How cool would it be to have something like this working for you while hiking in the wilds of India, or China, or any other Federation planet?

Of course, who knows how this device might fare in the field--machine language translators are faced with a daunting test, when posed with the sloppiness of real-life language. Just witness how bad Google's translation still is. The field has made tremendous strides in recent years. Will the pace continue?