Algorithm 1 The proposed algorithm

Input: The received signals

Step 1: Segment the received signals based on the known or estimated hop timings. Step 2: In one hop duration, calculate the STFT $X\left(t,f\right)=\left\{X_1\left(t,f\right),\dots ,X_M\left(t,f\right)\right\}$ of the received signals. For each TF point $\left(t_a,f_a\right)$, calculate ${\displaystyle \sum _{p=1}^{M-1}{\displaystyle \sum _{q=p+1}^M\left|\frac{\left|X_q\left(t_a,f_a\right)\right|}{\left|X_p\left(t_a,f_a\right)\right|}-1\right|}}$ and utilize Theorem (8) to detect signal-source-points. Step 3: Calculate $\frac{X_1\left(t_a,f_a\right)}{X_1\left(t_a,f_a\right)}$ for all SSPs. Based on (11), ${\mathsf{\Omega }}_{2,n}$ can be obtained and ${\widehat{\mathsf{\Omega }}}_{2,n}$ is calculated via clustering method. Step 4: Calculate the mixing matrix by (12). Step 5: Estimate $\widehat{\mathbf{S}}(t,f)$ by (13), and calculate the inverse STFT to obtain FH signals. The frequency $\{\widehat{f}, n=1,\dots , N\}$ can be calculated by FFT. Step 6: Calculate DOAs $\left\{{\widehat{\theta }}_n, n=1,\dots ,N\right\}$ by (14). Step 7: In another hop duration, repeat Step 2 to Step 6. Step 8: Compare the DOAs in different hop durations. The recovered FH signals can be spliced according to (15). Output: $\left\{{\widehat{\theta }}_n, n=1,\dots ,N\right\}$, $\left\{{\widehat{f}}_n, n=1,\dots ,N\right\}$ and recovered FH signals.