Abstract
In recent years, the advancement in voice-based authentication leads in the field of numerous forensic voice authentication technology. For verification, the speech reference model is collected from various open-source clusters. In this chapter, the primary focus is on automatic speech recognition (ASR) technique which stores and retrieves the data and processes them in a scalable manner. There are the various conventional techniques for speech recognition such as BWT, SVD, and MFCC, but for automatic speech recognition, the efficiency of these conventional recognition techniques degrade. So, to overcome this problem, the authors propose a speech recognition system using E-SVD, D3-MFCC, and dynamic time wrapping (DTW). The speech signal captures its important qualities while discarding the unimportant and distracting features using D3-MFCC.
TopIntroduction
In day-to-day life, people will frequently use textural, pictorial, and speech to share their information. But among those techniques speech is one of the powerful ways for communication. There are various studies is conducted on speech identification, verification, conversion (speech to text), and emotion detection Furui et al., 2004). The speech signals are classified into three different types based on the requirement of the study. Figure 1 represents the classification of speech signals.
Figure 1. The classification of the speech signal
Here, the speech signal is consists of countless messages and it is extracted based on the interest of speaker attention. The speech is the representation of waveform which is most useful in various practical applications. However, in this chapter, the author is mostly concentrated on speaker recognition systems (SRS). The purpose of SRS is to deploy this system in the forensic department which helps to identify the culprits.
Speech Recognition (SR)
In the era of computing, there are various speech processing techniques have been proposed by different researchers. The motive is to encourage high-speed development algorithms, computational architecture, and hardware. Speech recognition is the ability of a machine to identify the spoken words that are carried out in voice (Kushida et al., 2007). The digital words are in the form of a sequence that is matched against the coded dictionaries. Speech recognition is classified into two aspects (i) the system has to train for the identification of the patterns, and (ii) to identify the continuous or discrete word. The implementations of speech in the field of normal condition will become the alternative of keywords. However, the principle of speaker recognition is to identify and verify based on the speech wave (Sinha et al., 2019). This will help in many applications like voice dialing, speech-based banking, shopping, voice mail, security based on speech, forensic applications like determining the person's authenticity using the speaker verification process (etc.).
Speaker Verification (SV)
It is the process to identify the person whom it may claim to be. There is various literature found for speaker verification in terms of voice verification, authentication, speaker/talker authentication, and talker verification (Salman et al., 2007). It is a one-to-one comparison (a binary decision) between the input voice and claimed voice that is registered in the database. The basic classification of speaker identification is divided into three main components (i) Front-end Processing, (ii) Speaker Modeling, and (iii) Pattern Matching.
Figure 2. The basic classification of speaker verification
Figure 2 represents the speaker verification model. In this, the front-end processing is used to perform the pattern matching and it is based on the matching scores. If the score is larger than the threshold value, then the claimed speaker is recognized. Once the first phase is over then feature vectors of the speech signal are obtained and it is helped to detect the pattern while comparing it with the imposter database. If the match score value of the claimed person is high then the threshold value will produce high safety, on the other hand, it is also possible that a high degree of security will also lead to the rejection of the genuine person.
Key Terms in this Chapter
Feature Extraction: It is a process that starts with the initial set of measured data and builds derived values intended to be informative.
Preprocessing: It is used for cleaning, normalization, transformation, feature extraction, and selection. The product of data preprocessing is the final training set. Data pre-processing may affect how outcomes of the final data processing can be interpreted.
Speech Recognition: It is an interdisciplinary field of computer science and computational linguistics that develops methodologies and technologies for spoken language recognition.
Classifications: It is the process to identify the set of categories a new observation belongs to, based on a training set.
Approximation: It is the process where any data is intentionally similar but not exactly equal to something else.
Data Security: It is used to protect digital data, such as those in a database, from destructive forces and the unwanted actions of unauthorized users.
Mel-Frequency Wrapping: It is the purpose of grouping here is that each FFT value is multiplied against the corresponding filter gain and the result is summed.
Signal-to-Noise (SNR): It is the ratio that is used to measure the level of background noise.
Root Mean Square (RMS): It is a quadratic mean and is a particular case of the generalized mean with exponent 2.