Keyword recognition is one of the most significant problems in speech recognition systems. It has various applications in speech recognition, such as wake word detection, password detection, automatic transcription, speech detection, speaker identification, and many others. To the best of our knowledge, only a small amount of work has been published on text processing of the Dogri language and no work on speech processing has been published for the said language. In this paper, we present a robust approach for keyword recognition in audio files of Dogri Language using Gaussian Mixture Models (GMMs) in combination with Mel-Frequency Cepstral Coefficients (MFCCs). The proposed model effectively extracts audio features from audio files using MFCCs and utilizes probabilistic modeling such as GMM to classify spoken keywords. The dataset for the experiment is created in a noise-free environment with 2 male and 2 female speakers. To demonstrate the efficiency of the proposed approach, extensive experimentation is conducted on the created dataset and we obtained the micro avg. precision, recall, F1-score, and accuracy of 0.94, 0.96, 0.94, and 0.96, respectively.