Wave->Pitch Algorithm Selection

Since WIDI 4.0, settings of Wave->Pitch and Pitch->MIDI tabs are not separate. Every algorithm selected in Wave->Pitch effectively uses settings from the Pitch->MIDI tab as well, so we now call these algorithms "Note Recognition" algorithms rather than "Pitch Detection" or "Wave->Pitch" algorithms.



Euclid Algorithm

Euclid is a general purpose Note Recognition algorithm first included in WIDI 4.0. It features complex multi-pass recognition process including a lot of heuristics and fuzzy logic, and takes care of:

Overtone filtering
Note onset and end alignments
Filtering out random erroneous notes
Vibrato suppression

Generally Euclid with default parameters should be the first try when recognizing. This algorithm better fits moderately complex pieces with clear sounds. It is known to work very accurately with polyphonic piano or acoustic guitar compositions. However, this algorithm sometimes misses notes, especially in the bass range. If too many notes are missing, you may try to use Advanced Sensors instead.


Detection Sensitivity is a general control for setting how many notes will be in the recognized file. Set this parameter to lower values if too many erroneous notes are detected, and increase it if many notes are missing. You may also try to lower Overtone Filtering for decreasing number of missing notes.


Overtone Filtering enables you to prevent overtones to become a regular notes. Increase this parameter if you get chords instead of single notes, and decrease it if you hear less voices than you expected. Decrease this parameter (sometimes to a very low value) if many notes are missing.


Vibrato Suppression controls whether an attempt is made to suppress note fragmentation when vibrato playing technique is used. Note fragmentation is often heard as a "tremolo" sound where you expect to see only one note. However, setting this parameter to a very high value may "glue" real repeating notes with the same pitch.


Drum Detection

Drum detection is a special algorithm dedicated to recognition of percussion tracks. Percussion instruments can be detected with good time accuracy (<10ms), and their volumes can be determined. Percussions can be also divided into different instruments based on typical pitch. While this algorithm's purpose is to detect pure percussion tracks without any melodic instruments, it is sometimes possible to extract drums from a full composition as well.


Detection sensitivity is a general sensitivity control. Set this parameter to lower values if too many drum strokes are detected and increase it if many strokes are missing.


Ignore velocities sets all stroke velocities to the same value.


Differentiate Instruments by Pitch allows to determine average pitch of every drum stroke and assign strokes to different drums based on this value. Currently three pitch ranges are supported - bass, medium and high-pitched. You can set up margins and assign instruments for all three ranges.


Often high-pitched drums of the same MIDI velocity actually sound much quieter than bass ones. Boost parameter allows to compensate for that.


Advanced Sensors

This method is relatively fast, it uses a local auto tuning and harmonic filter algorithm. This algorithm was first introduced in WIDI 2.7. Peculiarities of this algorithm include large amount of extra notes (but it rarely completely misses notes), relatively inaccurate note starts and ends. However, it is worth a try, especially if the piece is very complex and you want to get just something useful, or if Euclid misses too much notes.


Local Auto Tuning adjusts fundamental frequencies of Sensors based on real spectral components of the source. It is useful when audio file you are transcribing is recorded with tones, shifted relative to the standard frequencies. Compared to setting Global Tuning Determination, this feature allows "local" tuning based only on the notes that are near current recognition position. Therefore, it is usually have no sense to use Global Tuning Determination except for informational purpose if Local Auto Tuning is set.


Harmonic Filter reduces the amount of  overtones that are recognized as separate notes.


Monophonic Mode detects only the loudest voice from a source file. It is recommended to set this parameter when recognizing one-voice melodies.


Time Shift of Bass Notes corrects starting times of bass notes compared to trebles. Sensors usually pushes bass notes further in time, so increasing of this parameter shifts them back. So if you hear that bass voices are sounding after trebles, try to change this parameter. Reasonable values are usually 0..40% of the scale.




WIDISOFT | MP3 and MIDI Software | Widi Recognition System | Able MIDI Editor | WIDI Audio To MIDI VST
Download | Mp3 to MIDI Samples | Presets | Order
MP3 to MIDI in detail | what is MIDI?