Multi-Device Hotword Detection Using Loudness-Based Response Delay

Each computing device calculates a delay value based on its loudness score before actually processing the hotword utterance. This preliminary delay calculation ensures that only the nearest device (with highest loudness score) will process the utterance first, while other devices are already delayed and will not process it. This resolves the contradiction by maintaining spatial flexibility through multiple monitoring devices while eliminating redundant processing and reducing latency through pre-calculated delays.

The patent implements local quality by making each computing device's processing behavior dependent on its local loudness score relative to other devices. The device with the highest loudness score (nearest to the user) processes the utterance, while other devices with lower scores are suppressed through delay mechanisms. This ensures reliable utterance detection by the most appropriate device while conserving energy by preventing redundant processing across all devices.

The patent changes the parameter of loudness score to discriminate between intended and unintended utterances. By calculating loudness scores for each device and using these scores to determine processing priority, the system can distinguish user-directed hotwords from background noise and unintended speech. The delay mechanism uses these parameter changes to ensure only the device with the highest loudness score processes the utterance, resolving the contradiction between environmental coverage and utterance discrimination.