Methods and apparatus are provided for improving wake-up word (or trigger word) detection by an audio device. After initially detecting a WUW, an audio device validates the detected WUW using inputs from one or more other systems such as voice activity detection (VAD), on-head detection, or other headphone data. Other headphone data includes inputs received via sensors on the audio device that provide contextual information associated with a state of the user. Based on the inputs from other systems, the audio device is able to identify unintended WUW activations and increase WUW detection accuracy.