Sloud QBH Search

query by humming


Products & Technologies

Corporate Info

Sloud Search music by humming (QBH)

Imagine you are listening to a musical theme and fall in love with it. You want to hear it again. Or perhaps you want to buy the record of that tune that your beloved loves to hear. But you don't know the title or the composer of this particular piece. You start asking your friends, humming the melody. To your disappointment nobody can help you. So you are stuck. Years go by and you still once in a while think about that mysterious tune you can never figure out what it is... Well... you are not alone. This problem is a very old one...
-- from Niclas Fogwall, The search for a notation index.

Sloud Query by Humming (the QbH ActiveX applet is on the front page of the site) is a search technology that allows you to find music by singing a clip in your own voice. Sloud uses fuzzy search algorithms that allow for a certain amount of imprecision and mistakes in the singing.

Search depends on your ability to sing. Having musical ears and an ability to sing well would greately improve the search results. Sloud search is similar to the usual text searches: if one enters the query string with typos, the search engine will find pages with the same typos, not the intended correct words. Likewise, Sloud will find matches to what you actually sang, not what you meant to sing.

For the best results please sing "DA-DA-DA", do not sing the lyrics. Try not to sing any other words but DA-DA-DA.

Notice the indicator window. While singing, you will see colored stripes appearing in the window. If the stripes are not appearing, check weather the microphone is attached and working properly. If the microphone works but the stripes are still missing, sing louder.

The colors of the stripes indicate the recognized tones. Try to keep stripes uniformly colored. More uniform color of the stripe means the tone is identified better.

Colors of the stripes

Press the Record button and sing one piece to the microphone. Press the Stop button. If you do not press the Stop, the recording will be automatically terminated in about 30 seconds.

The music fragment that you are using as a key for the search should not be shorter than seven notes. For better results use longer and more characteristic fragment of the melody.

Try to sing carefully indicating all notes. Do not try to sing rapid music in its real tempo. Missing notes degrade search results more than slowing down the tempo.

Sing the music in a pitch that is convenient for your voice. Do not try to sing close to the original pitch if you find it difficult, for example there is no need to sing too low or too high.

It is helpful to listen to the recorded fragment before sending it to the search program. If by listening you cannot recognize the clip that you created, Sloud Search will also likely have difficulty recognizing it. In such a case you may want to sing it again.

Advanced Search

When the "Exactly" box is checked, Sloud takes into account both the rhythm and the pitch of the music. If the option is turned off, Sloud completely disregards the rhythmical movement of the melody. That may produce better search results in cases when you do not clearly remember the music.

You may improve search results by limiting it to a subset of all music. For example, if you know the name of the artist/band, enter it in the appropriate field. The search will return records only from the given artist or band.

Search by words in the Title or Lyrics

Use one or more words which are essential for the search result. The accuracy of the searches will be better if several words or a phrase fragment are submitted.

Sloud Search by MIDI Tune fragment

This feature allows to search by a fragment of music such as a tune or a theme. The tune fragment must be represented by a MIDI file that includes the sequence of notes you search for. This MIDI file should contain only one tune presented by 7-30 notes in a single MIDI channel. The chords will terminate the search.

For More Information:

Download article "Sloud Query-by-Humming Search Music Engine" (PDF format)


 

 
© 2005-2007 Sloud Inc.