Summary
In space‐based automatic identification system, message collision is inevitable. To separate collision messages in multichannel space‐based automatic identification system, an improved least squares constant modulus algorithm (LSCMA) is proposed in this paper. The algorithm initializes LSCMA by using the result of complex symmetric fast independent component analysis and reduces the dimension of the received signal by using the minimum description length algorithm. At the same time, a new metric named failure rate is proposed to measure the separation performance. The simulation shows that the performance loss caused by the delay, the frequency and phase offset, and the input signal‐to‐interference ratio is within an acceptable range. The proposed algorithm outperforms the original LSCMA and other algorithms in terms of red packet error rate and failure rate.