Radar-based fall detection technology has attracted much attention for its high accuracy, robustness, and privacy preservation potential. Detection of “Soft Fall”, i.e., high-freedom fall, is the key to practical application. This paper proposes a novel height-tracking method based on a multiple-input multiple-output (MIMO) radar system to address this problem. First, the received signal was segmented into a time sequence with a sliding window along slow time. Next, Fast Fourier Transform (FFT) and Multiple Signal Classification (MUSIC) algorithms were applied to estimate the general trend of the human body’s time-varying range and pitch angle information. Then, they were fused with a geometrical relationship to describe height changes during fall motions using a height trajectory map. Two height-based features were extracted as input to Support Vector Machine (SVM) to distinguish soft fall and fall-similar motions. Finally, experiments, including four soft fall and five typical fall-similar motions, were conducted to demonstrate its feasibility and superiority.

This paper describes the multi-query multi-head attention (MQMHA) pooling and inter-topK penalty methods which were first proposed in our submitted system description for VoxCeleb speaker recognition challenge (VoxSRC) 2021. Most multi-head attention pooling mechanisms either attend to the whole feature through multiple heads or attend to several split parts of the whole feature. Our proposed MQMHA combines both these two mechanisms and gain more diversified information. The margin-based softmax loss functions are commonly adopted to obtain discriminative speaker representations. To further enhance the inter-class discriminability, we propose a method that adds an extra inter-topK penalty on some confused speakers. By adopting both the MQMHA and inter-topK penalty, we achieved state-of-the-art performance in all of the public VoxCeleb test sets.