The human voice is a perfect source of data for person identification in many applications. With a growing need for security in various public places, voice biometrics seems to be a good solution, since voice records can be easily taken. This paper provides a brief overview of approaches used in speaker recognition. It also presents a novel approach for speaker recognition in degraded conditions in the context of smart homes. The proposed method includes a pre-processing phase, feature extraction phase and classification phase. The second phase is based on the extraction of the sound spectrum energy maxima of the speech sound called formant. The next step is the use of a well-known technique called Dynamic Time Warping to find an optimal alignment between two given (time-dependent) sequences under certain restrictions. Complimentary steps are needed to complete the processing. The experiments are performed on a 1248-sample database to validate the proposed approach. The novel approach leads to better results regarding the state-of-the-art with an accuracy of 94.5%.