Analog systems and methods for audio feature extraction and natural language processing

An all-analog natural language processing method, comprising:

• receiving an analog audio signal from an analog microphone at an analog filter bank;
• filtering the analog audio signal, by a plurality of analog bandpass filters configured for detecting voice activity, to extract features;
• classifying extracted features, by a voice activity analog neural network to generate an output of voice features;
• and
• classifying the extracted features, by a word detection analog neural network to generate an output of a keyword or a wake word.

A phase-based audio buffer, comprising:

• a dual-slope voltage to phase converter configured to convert an audio signal voltage received from an analog microphone into a phase representation;
• a first counter configured to digitize the phase representation for storage in a first-in-first-out sub-threshold memory structure;
• a second counter configured to convert the phase representation stored in the memory structure to time pulse widths;
• a comparator for comparing the time pulse widths against an arbitrary threshold to output an analog pulse using one pin;
• and
• a gated ring oscillator as a time base to convert the audio signal voltage into the phase representation and convert the phase representation into the time pulse widths.