This paper proposes a new radar tracking filter named Noise-estimate Particle PHD Filter (NP-PHDF). Kalman filter and particle filter are popular filtering techniques for target tracking. However, the tracking performance of the Kalman filter severely depends on the setting of several parameters such as system noise and observation noise. It is an open problem how to choose proper parameters for various scenarios, and they are often regulated in trial-and-error manner. To tackle this problem, Noise-estimate Particle Filter (NPF) has been proposed so far. The NPF estimates proper noise parameters of a Kalman filter on-line based on a scheme of particle filter. In this paper, we extend the NPF so that it enables to track multiple targets simultaneously by combining with Probability Hypothesis Density (PHD) filter, and propose a new Noise-estimate Particle PHD Filter (NP-PHDF). Simulation results show that the proposed filter has higher tracking performance in various scenarios than conventional Kalman filter, particle filter, and PHD filter for multiple-targets tracking.