New features added to MIRtoolbox 2 as part of the MiningSuite
Audio/symbolic articulation:
- MIDI files can be loaded and represented using same syntax.
- Symbolic and audio representations can both be analysed using same methods (for tempo and tonal analysis for instance).
- Audio can be transcribed into symbolic representation and further analysed based on new methods.
Speed-up:
- More efficient computation using less RAM.
- No more display of all the successive steps in the Command Window. Speeds up the computation.
- A command terminated by a semi-colon (
;) is not executed at all. This enables to build a series of operations efficiently. In this way, the'Design'keyword from MIRtoolbox is not necessary anymore.
Input/output:
- You can import your own data and represent as an object of class available in the MiningSuite (for instance
sig.Signalorsig.Spectrum). - Each result also stores the complete description of the list of operations with all the specified options and parameters.
Multichannel:
- Multichannel audio input are not summed together by default, but this summation can be performed in a first step by calling
'Mix'option on any operator. - New type of graphical representation of multichannel data.
- There can be several layers of channel decomposition in a given signal. For instance, a stereo waveform can be further decomposed into bands using
sig.filterbank. -
sig.filterbank(..., 'Frame')performs the frame decomposition after the filterbank decomposition.
Additional layer of syntax in the code:
- Matrices imported into, used in, and exported from the MiningSuite have their internal structure clarified: the role of each dimension is made explicit using a systematic formalism provided by the
sig.dataclass. All numerical data is represented as objects of thesig.dataclass. - Architecture of the toolbox much simpler, allowing faster computation and more transparent and clear code.
Other new features:
sig.audio: 'Detrend' option
Back to Transition from MIRtoolbox
