The growth in the number of low-cost narrow band radios such as Bluetooth low energy (BLE) enabled applications such as asset tracking, human behavior monitoring, and keyless entry. The accurate range estimation is a must in such applications. Phase-based ranging has recently gained momentum due to its high accuracy in multipath environment compared to traditional schemes such as ranging based on received signal strength. The phase-based ranging requires tone exchange on multiple frequencies on a uniformly sampled frequency grid. Such tone exchange may not be possible due to some missing tones, e.g., reserved advertisement channels. Furthermore, the IQ values at a given tone may be distorted by interference. In this paper, we proposed two phase-based ranging schemes which deal with the missing/interfered tones. We compare the performance and complexity of the proposed schemes using simulations, complexity analysis, and measurement setups. In particular, we show that for small number of missing/interfered tones, the proposed system based on employing a trained neural network (NN) performs very close to a reference ranging system where there is no missing/interference tones. Interestingly, this high performance is at the cost of negligible additional computational complexity and up to 60.5 Kbytes of additional required memory compared to the reference system, making it an attractive solution for ranging using hardware-limited radios such as BLE.

翻译：随着低功耗蓝牙（BLE）等低成本窄带无线电在资产追踪、人类行为监测和无钥匙进入等应用中的普及，精确的距离估计成为此类系统的关键需求。相较于传统基于接收信号强度的测距方案，基于相位的测距由于在多径环境中具有高精度而受到广泛关注。然而，该方法要求在均匀采样频率网格上进行多频点音调交换，但实际应用中常因预留广播信道等机制导致部分音调缺失。此外，特定音调上的IQ值可能受到干扰影响。本文提出两种处理缺失/干扰音调的相位测距方案，通过仿真实验、复杂度分析和实测系统评估其性能与复杂度对比。研究表明，当缺失/干扰音调数量较少时，基于训练神经网络（NN）的方案与无缺失/干扰音调的参考测距系统性能高度接近。值得注意的是，相较于参考系统，该方案仅增加极低的计算复杂度和最多60.5 KB的额外存储需求，使其成为适用于BLE等硬件受限无线电测距的理想解决方案。