In this paper, two new Cooperative Kalman-Bucy filters are derived using the matrix theoretic consensus based approach to cooperative control. The first is a centralized design of Cooperative Kalman-Bucy filter for which stabilizing and optimal gains are found to minimize error state covariance. Further, a distributed design of Cooperative Kalman-Bucy filter is also proposed by explicitly accounting for available information. Both designs of the proposed new Cooperative Kalman-Bucy filter can be applied to a team of heterogeneous time-varying systems within an incomplete, unidirectional communications network, and the overall estimation performance is superior to individual Kalman filters as long as better sensors are located at globally reachable nodes.