Experiment 4: Fixed energy band vs. formant tracking
Experiment 4: Fixed energy band vs. formant tracking
Hypothesis
- Fixed energy band can perform as well as formant tracking
Assumptions
- Performance does not address insertion errors
- Measure performance by vowel "detections"
- Also observe violations of theoretical prediction
Methodology
- Redo experiment 1 using fixed energy band
- Attempt to optimize band edges and transition widths
Results
- Canonical values (300 to 900 Hz) do as well as formant tracking
- Performance is not sensitive to changes in these values