In Figure 5.14, we see results of MCMCUMS, similar to Figure 5.11. Rather than using synthetic data, we captured a segment of 768 samples of human speech at 12kHz, then calculated the MFCC features . With this feature value fixed, we used MCMCUMS to produce random spectra. On the top graph, we see one random MCMCUMS sample. On the bottom, we show the LP solution, the MaxEnt solution , and the average of 10000 full MCMCUMS iterations. Once again we can conclude that precisely predicts the mean of the MCMCUMS generated spectra. Important to note is that is a very smooth spectral estimate that is, at least visually, very satisfactory. Our proposed method may be preferred to openended MFCC synthesis methods [39] since the resulting spectrum is featurereproducing and has optimal smoothness as a result of satisfying the maximum entropy rule. This is because maximizing the entropy (5.14) is the same as maximizing the spectral flatness [36,25].

Baggenstoss 20170519