We investigate the asynchronous multiple user access communication in optical wireless scattering communication, where different users transmit signals without perfect alignment in the time domain. Firstly, we characterize the received signal based on hidden markov model (HMM) such that the misalignment among different users can be characterized by the state transition. Then, we investigate the achievable rates based on that of the HMM and obtain the approximated solution using Monte Carlo method. We propose the channel estimation based on expectation-maximization (EM) algorithm. Furthermore, we adopt Viterbi and Bahl-Cocke-Jelinek-Raviv (BCJR) algorithms for joint iterative multi-user decoding. Numerical and experimental results illustrate the performance of proposed channel estimation, joint detection and decoding. It is seen from the experimental results that the proposed approaches perform close to the simulation results.