Feature Set 1: MFCCs
Feature Set 2: MFCCs + Correlation Dimension
Feature Set 3: MFCCs + Lyapunov Exponent
Feature Set 4: MFCCs + Correlation Entropy
Feature Set 5: MFCCs + Correlation Dimension + Lyapunov Exponent + Correlation Entropy

Test Set 1: Clean
Test Set 2: Airport Noise Added
Test Set 3: Babble Noise Added
Test Set 4: Car Noise Added
Test Set 5: Restaurant Noise Added
Test Set 6: Street Noise Added
Test Set 7: Train Noise Added

Evaluation Results for Aurora Corpus (WER)
	Test Set 1	Test Set 2	Test Set 3	Test Set 4	Test Set 5	Test Set 6	Test Set 7
Feature Set 1	13.5%	53.0%	55.9%	57.3%	53.4%	61.5%	66.1%
Feature Set 2	12.2%	57.1%	59.1%	65.8%	55.7%	66.3%	69.6%
Feature Set 3	12.5%	56.8%	60.8%	60.5%	58.0%	66.7%	69.0%
Feature Set 4	12.0%	52.8%	56.8%	58.8%	52.7%	63.1%	65.7%
Feature Set 5	12.8%	58.6%	63.3%	72.5%	60.6%	70.8%	72.5%

The following results use feature sets similar to those above, but with the adition of the first and second derivative of the invariants.

Feature Set 6: MFCCs + Correlation Dimension + 1st and 2nd Invariant Deltas
Feature Set 7: MFCCs + Lyapunov Exponent + 1st and 2nd Invariant Deltas
Feature Set 8: MFCCs + Correlation Entropy + 1st and 2nd Invariant Deltas

Evaluation Results for Aurora Corpus (WER)
	Test Set 1	Test Set 2	Test Set 3	Test Set 4	Test Set 5	Test Set 6	Test Set 7
Feature Set 6	12.3%	61.4%	66.1%	68.6%	60.4%	73.8%	75.7%
Feature Set 7	12.0%	58.2%	60.8%	60.8%	60.2%	68.0%	68.8%
Feature Set 8	12.1%	56.3%	59.8%	60.6%	55.1%	65.0%	68.8%