End-to-end Alexa Device Arbitration
This work addresses a practical problem for smart home systems by enabling more accurate device selection, though it is incremental as it builds on existing speaker localization concepts.
The paper tackles the device arbitration problem, determining which smart home device is closest to a user based on keyword detection, by proposing an end-to-end machine learning system that learns feature embeddings from each device and aggregates them for decision-making, achieving results compared against a signal processing baseline.
We introduce a variant of the speaker localization problem, which we call device arbitration. In the device arbitration problem, a user utters a keyword that is detected by multiple distributed microphone arrays (smart home devices), and we want to determine which device was closest to the user. Rather than solving the full localization problem, we propose an end-to-end machine learning system. This system learns a feature embedding that is computed independently on each device. The embeddings from each device are then aggregated together to produce the final arbitration decision. We use a large-scale room simulation to generate training and evaluation data, and compare our system against a signal processing baseline.