/ A Timbre Analysis And Classification Toolkit For Pure Data
ï~~4.3. Timbre ordering The timbre ordering examples use two different approaches to sound segmentation: the first patch reads in pre-determined onset/offset times for each of 51 percussion instrument attacks, and the second automatically divides loaded samples into grains that are 4096 samples in length by default. Onset/offset labels for the first example were generated manually in Audacity, exported to a text file, then imported to a table in Pd. The percussive sound set included with this example is small, and is intended to provide a clear demonstration of timbrelD's ordering capabilities. Figure 5 shows a region of the patch that includes the table where ordering information is stored and 5 sliders that control feature weighting. ard inr $1 -- m Figure 5. 51 percussion sounds ordered based on a userspecified weighting of 5 features. Ordering is always performed relative to a user-specified starting point. With 51 instruments, when an instrument index between 0 and 50 is supplied along with the "order" message, timbrelD will output the ordering list at its fourth outlet for graphing. Using the 5 feature weight sliders, it is possible to boost or cut back the influence of any particular feature in the ordering process. The features implemented in this patch are temporally evolving spectral centroid, spectral flatness, zero crossing rate, loudness, and BFCCs. After hearing the results of a particular ordering, the levels of the feature weight sliders can be changed in order to produce a new ordering and gain an understanding of the effects of various features in the process. An ordering is shown in the graph of Figure 5, where the y axis represents instrument indices 0 through 50, and the x axis indicates each instrument's position in the ordering. It begins at in strument 0 with a drum and progresses through other drum strikes followed by snares, a sequence of cymbal strikes, and a sequence of wooden instruments. Ordering the set by starting with a wooden instrument will produce a different result that retains similarly grouped sequences. An expanded version of this patch could be useful as a compositional aid for exploring relationships between sounds in a much larger set, offering paths through the sounds that are smooth with respect to different sonic characteristics. Two types of ordering are available: "raw" and "relative". The graph in Figure 5 was produced with relative ordering, which starts with the user-specified instrument, finds the nearest match in the set, then finds the nearest match to that match (without replacement), and so on. The point of reference is always shifting. Raw ordering begins with the given instrument, then finds the closest match, the second closest match, the third closest match (also without replacement), and so on. Orderings of this type start with a sequence of very similar sounds that slowly degrade into randomness, and usually finish with a sequence of similar sounds-those that are all roughly equal in distance from the initial sound, and hence, roughly similar to each other. The second ordering example loads and segments arbitrary sound files. Loading a speech sample generates sequences of similar phonemes with a surprisingly continuous pitch contour. Audio generated from this and other ordering examples can be accessed at the author's website. 4.4. Mapping sounds in timbre space Figure 6. 847 speech grains mapped with respect to the 2nd and 3rd BFCC. Another way to understand how the components of a sound set relate to one another is to plot them in a userdefined timbre space. CataRT is the most recognized and well developed system for this task; timbrelD makes it pos 228
Top of page Top of page