COVID-19 is a viral pneumonia that causes symptoms in the lungs of infected individuals. The presence of the symptoms must be diagnosed as soon as possible.Other than RTPCR test, One of the most common diagnosis for any lung related infection is by having an X-ray. In most cases, the goal is to differentiate between healthy individuals, viral Pneumonia and Covid-19 cases. Lung infection diagnosis can be performed with computeraided diagnosis of a patient’s chest X-ray scan for a quick and accurate diagnosis. In view of having a more accurate automated system, a hybrid transfer learning method with Error-Correction Output Codes (ECOC) was proposed to enhance the automated diagnosis. The proposal first considers the frozen features existing network without any training. This serves to preserve generalization. Subsequently, the features were concatenated from a feature vector. However, instead of implementing the features in a single multi-class single-machine learning model, an ensemble of machine learning methods was proposed. In particular, the ensemble Error Correction Output Code (ECOC) was considered. By combining network features including GoogLeNet, ResNet-18, and ShuffleNet for feature extraction, the results were tested against the conventional fine-tuning approach of Transfer Learning (TL). X-ray input data were collected from the open-source repositories. In this implementations, Support Vector Machine (SVM) as the base classifier. The proposed network attempts to categorize the input data into one of three categories: COVID-19, healthy, or non-COVID-19 pneumonia. The mean accuracy of our method was 96.21% compared to the existing fine-tuning pre-trained model, which yielded 89.1% for Goog-LeNet, 88.95% for ResNet-18, and 89.31% for ShuffleNet. This strongly suggests that an improvement is achieved owing to the inclusion of features from various networks and a more complex final classification layer, which is the ensemble configuration.