A review on speaker recognition: Technology and challenges

Fig. 2. Framework of a speaker identification system.
Table 3
Comparison of different feature extraction techniques.

Technique	Merits	De-merits
LPC	Based on basic principles of sound production [20]. Simple to implement and mathematically precise [19].	Performance degradation in the presence of noise [20, 23]. It does not represent vocal tract characteristics from the glottal dynamic; thus, consumes time and computational cost [9]. Inconsistency with human hearing [24].
LPCC	Smoother spectral envelope and stable representation as compared with LPC [20]. Feature components are decorrelated [19, 25].	Linearly spaced frequency bands, which is inadequate [19,25]. Sensitive to the quantization noise [23]. The performance is degraded in case of using insufficient order [23].
MFCC	More information on lower frequencies than higher frequencies due to mel-spaced filter banks. Hence, behaves like human ear compared with other techniques [20, 26]. Captures the main characteristics of phones in speeches with low complexity [27].	Based on Short-Time Fourier Transform (STFT) which has fixed time-frequency resolution [26]. Low robustness to noise [19, 28].

1. 
   Authentication
   

Authentication is one of the most popular biometric applications as it allows the users to identify an individual based on his/her voice. Usually, to authenticate the speaker, a combination of techniques is used, such as a password or facial recognition [22]. This biometric authentication could reduce the problems of misused identity and is also more convenient than using a Personal Identifi­cation Number (PIN) or password, which can be easily forgotten. In 2015, Gomar invented and patented a speaker recognition system for authenticating a mobile device user [29]. The system produced a biometric voiceprint from the user’s speech utterances which are stored in the mobile device when it met a quality threshold. For verification purposes, the user will need to utter an attribute from the biometric voiceprint in order to be an authorized user.

2. 
   Personalization
   

A personal digital assistant used to be a luxury reserved for the wealthy, but it is no longer the case nowadays as everyone can have a digital assistant. Siri, Alexa, Cortana, and Google Now, to name just a few, are examples of software that can help us comprehend and carry out our capricious spoken commands. We can do anything with them, such as planning for meetings, event scheduling, shopping,