Automatic modulation identification is a process to identify the modulation type of received signal by observing its features. It has a vital role in many communication systems in both the civilian sector and the military sector. The received signal is usually corrupted by the effect of different sources like Gaussian noise, fading and interference, which impact on the signal quality. Due to the extensive usage of digital modulation signals in different applications, most recent researches are focused on recognizing these type of signals. A wide variety of methods is utilized for determining the modulation type, and these methods can be categorized into two main categories: decision-theoretic and feature- based. In the most existing researches, only the effect of Gaussian noise and fading have been taken into account, and the interference effect has been ignored. In this research, the performance of the classifier in the co-channel interference scenario was addressed, but when the classifier has been trained solely on interference-free data, using a feature- based approach. Moreover, the effect of carrier frequency mismatch, symbol rate deviation and a number of samples in the signal were investigated. Two feature extraction methods are used in this project: instantaneous time-domain features method and cumulant features method with an artificial neural network (ANN) classifier. These feature extraction methods are chosen based on their advantages. In addition, for the cumulant features method; two models are used using different key features in each model. The classification accuracy shows that the cumulant feature outperforms the instantaneous time feature. It was able to identify the desired class from interfering class at lower SIR (10 dB) with an accuracy of 85.3 % for model one and 48.7% for model two. Also, the results show that the classification accuracy affected by the aspects related to carrier frequency mismatch, symbol rate deviation and a number of samples in the signal. Finally, the advantage of the proposed system is investigated in terms of accuracy and design complexity compared to other works in the literature.