Metric Learning for Keyword Spotting
This addresses the issue of high false alarm rates in keyword spotting systems for real-world applications, though it is incremental as it builds on metric learning with a hybrid approach.
The paper tackles the problem of keyword spotting as a detection task, where existing classifier-based methods fail on unseen non-target sounds, leading to high false alarm rates. The proposed metric learning method significantly reduces false alarms to unseen non-target keywords while maintaining overall classification accuracy, as demonstrated on the Google Speech Commands dataset.
The goal of this work is to train effective representations for keyword spotting via metric learning. Most existing works address keyword spotting as a closed-set classification problem, where both target and non-target keywords are predefined. Therefore, prevailing classifier-based keyword spotting systems perform poorly on non-target sounds which are unseen during the training stage, causing high false alarm rates in real-world scenarios. In reality, keyword spotting is a detection problem where predefined target keywords are detected from a variety of unknown sounds. This shares many similarities to metric learning problems in that the unseen and unknown non-target sounds must be clearly differentiated from the target keywords. However, a key difference is that the target keywords are known and predefined. To this end, we propose a new method based on metric learning that maximises the distance between target and non-target keywords, but also learns per-class weights for target keywords à la classification objectives. Experiments on the Google Speech Commands dataset show that our method significantly reduces false alarms to unseen non-target keywords, while maintaining the overall classification accuracy.