Language Identification (LID) in Natural Language Processing (NLP) is the process of identifying the spoken language in speech utterances. In the last decade, the interest and functional application of speech processing systems have grown exponentially. The proliferated use of handsfree voice-operated devices, speech-to-speech translation systems requires low latency, reliable automatic speech identification systems. This article examines three different models to recognize languages automatically in speech. The first model uses Dynamic Hidden Markov Networks (DHMNet) for LID in utterances. Another model utilizes Deep Neural Network (DNN), and the third uses the recently developed Long Short-Term Memory (LSTM) Recurrent Neural Network (RNN). Finally, comparing three different models, it is shown that a fusion of LSTM RNN and DNN model gives better results than the state-of-the-art models when applied to short utterances.