Consumers have become more comfortable using voice to interact with technology as this interface has become common place on higher-end electronics device such as mobile phones and home automation appliances. While convenient and commonplace, voice recognition historically is high-power in the 10s of milliamp range. This session describes in details the technical process and engineering tradeoffs of adding voice commands to battery-powered fitness, medical and wearable products. This session will cover energy aware embedded system design techniques, technologies and statistical Markov model algorithms that reduce the total signal chain current consumption in order of magnitude into the micro-amps enabling always-on, hands-free voice command recognition.
