Determination of effective visual response time window is premise of studying object cognition and decision-making mechanisms in avian object-oriented task. However, it is difficult to determine the effective response time window for freely moving birds. In this study, video data of pigeon’s face in the target-oriented task was collected, and a neural network algorithm based on Faster RCNN was applied to train pigeon face prediction model, which was further used to predict the time window of pigeon observing the specific image target. The length of time window were estimated based on the statistical results of 177 trials, and the specific time window for each trial was then determined by combining the start frame that contained pigeon’s frontal faces. Finally, the proposed method was verified using data from two pigeons trained for object-oriented task. Taking the firing rates feature population recorded from pigeon’s ectostriatum as an example, the mean firing rate during the estimated effective time window and the original mean firing rate without the time window were sent to SVM and KNN classifier respectively to decode the observed object category. The comparison results showed that the classification accuracy of both classifiers were significantly improved with our method, proving that the proposed method could obtained effective response to specific visual object for freely moving birds.