利用粒子滤波重建位置细胞编码的运动轨迹

Decoding Movement Trajectory of Hippocampal Place Cells by Particle Filter

粒子滤波解码算法在神经信息解码中已有较多应用,但在海马区位置细胞集群编码的运动轨迹重建中极其少见.针对大鼠海马区位置细胞的神经元响应特性,采用二次指数泊松方程建立了大鼠运动轨迹的位置细胞集群状态空间编码模型,然后利用仿真数据和实测数据研究了粒子滤波在大鼠运动轨迹重建中的性能,并与扩展卡尔曼和无迹卡尔曼重建算法进行了对比.仿真数据重建结果显示,与后两种算法相比,在相同的重建精度下,粒子滤波算法需要的位置细胞个数相对更少.实测数据重建结果显示,粒子滤波算法重建的轨迹与真实轨迹之间的相关系数和均方根误差均优于扩展卡尔曼和无迹卡尔曼重建算法.这些结果表明,粒子滤波算法不仅能够高效地利用位置细胞集群编码信息,而且具有更高精度的轨迹重建性能,将为空间认知神经机制的深入研究提供有力的技术支持.

Particle filter（PF） algorithm has been applied in neural decoding,but rarely in movement trajectory reconstruction of hippocampal place cells.According to the response characteristics of place cells,the state space population encoding model of movement trajectory was established with quadratic exponential Poisson＇s equation in this paper,the performance of the PF algorithm was investigated with respective simulated data and real data in movement trajectory reconstruction.And the results were then compared with extended Kalman filter（EKF） and unscented Kalman filter（UKF） algorithms.For the simulated data,the number of place cells needed by the PF is less than that of the others,under the same reconstruction precision.For the real data,correlation coefficient and root mean square error between true trajectories and reconstructed trajectories by the PF are superior to that of by the EKF algorithm and by the UKF algorithm.These results demonstrate that not only does the PF algorithm efficiently utilize encoding information of place cells population,but also has an outstanding movement trajectory reconstruction performance.It would provide powerful technique support for further research in spatial cognitive mechanism.