← Back to the chapter · Previous detail · Next detail →
Technical detail · Audio Feature Extraction for IoT
Train a keyword model and validate each task
Those coefficients then have to be turned into a trained model, and the loop that does it is where the acoustic assumptions are quietly baked in. Work through Figure 7 from the top of the batch loop to the four heads at the bottom.
Note where Augment the clip sits in Figure 7: the waveform-level background-noise, time-shift, and volume changes shown come before Extract MFCC 98 x 13. Other training methods can augment extracted features, such as SpecAugment’s time and frequency masking of log-mel spectrograms; the deployed feature extractor still uses the recorded unaugmented inference recipe. The loop then closes through Cross-entropy loss and Adam or SGD update and repeats, which is why the split and the background-noise set are part of the model’s evidence rather than a preprocessing footnote. The lower card carries the discipline that multi-task audio work most often skips: shared hidden layers may serve keyword, speaker, stress, and environment at once, but each head needs its own label source, its own metric, and its own stated failure boundary. A trunk that is well validated for keyword detection has proved nothing about the speaker head hanging off the same layers.