Widespread use of English in the academia and in business is leading an increasing number of people to learn it as a second or a foreign language. Computer aided pronunciation training (CAPT) systems are used by non-native English speakers for improving their English pronunciation. A typical CAPT tool records the speech of a learner, detects and diagnoses mispronunciations in it, and suggests a way for correcting them. We classified the CAPT systems for English into four categories on the basis of the technology used in them and studied the salient features of each such category. We observed that visual simulation based systems are suitable for young and naive learners, game based systems are advantageous as they can be personalized as per the requirements of the learners, comparative phonetics based systems are suitable for adult learners fluent in another language, and artificial neural network based systems have the highest accuracy in mispronunciation diagnosis and are suitable for experienced and professional learners. We identified the state-of-the-art practices used in CAPT systems, and observed that CAPT systems can detect up to 86% mispronunciations in a speech and help learners to lessen mispronouncing by up to 23%. We recommend collaboration between language teachers and software developers to develop CAPT tools, their wide dissemination and integration with the curriculum at school and university levels, and further investigation on mobile and collaborative CAPT systems.