Comparison of MFCC and LPCC for a fixed phrase speaker verification system, time complexity and failure analysis

Abstract

This paper reports some of the observations perceived on uncontrolled environment database for comparison of the Mel Frequency Cepstral Coefficients(MFCC) and Linear Predictive Cepstral Coefficients (LPCC) for development of a robust fixed phrase speaker verification system. MFCC are Cepstral coefficients computed on a warped frequency scale based on known human auditory perception whereas LPCC are Cepstral coefficients that represents the human articulatory system based on linear prediction. This paper compares the accuracy level of both the systems based on MFCC and LPCC, also it compares the systems from an equal error rate point of view (EER). The result suggests that LPCC performs more accurately as compared to that of MFCC by 2.59% for authenticating a speaker. The study also suggests that on basis average time required for giving a decision, MFCC outperforms LPCC significantly by 3.73 sec. Furthermore, the paper includes an analysis for the failure of the system on some of the tests which are not correctly detected as genuine or imposter.

Detection of Snore from OSAHS Patients Based on Deep Learning

“…As shown in Table 5 and Figure 6 , through longitudinal comparison, it is found that the feature extraction effect from the best to the worst are MFCC, LPMFCC, LPCC. This is because the LPCC algorithm has a linear prediction function for time series and can obtain more information from speech recognition [ 28 ], while MFCC maps the speech frequency to a nonlinear mel filter bank and converts it to the cepstrum domain [ 29 ]. The features extracted from the neighboring frames are almost independent and suitable for consonant recognition.…”

Robust cepstral feature for bird sound classification

“…This feature is defined as the inverse Fourier transform (IFT) of the logarithmic magnitude of the linear prediction spectral complete envelope [30], and it provides a more robust and compact representation, which is especially useful for automatic speech recognition and speaker identification [25]. LPCC represents [31].…”

“…Audio signal processing researchers have proven that despite the volume of works that have been done using MFCC features for classification purpose, a spectrum of improvement in classification accuracy can still be achieved by considering other cepstral features such as GTCC [23], [24] and LPCC [25], [26]. Especially for non-speech audio classification [11] and noisy environmental audio data, GTCCs have been shown to be more robust [9].…”

Multidimensional zero-crossing interval points: a low sampling rate acoustic fingerprint recognition method

“…MFCC creates a sound feature extraction model according to human ear auditory characteristics [3]. Although LPCC and MFCC both have high recognition precision, the sampling rate is always over 8 kHz, resulting in large data amounts and complex computations, imposing high requirements for the hardware [4]. Therefore, the two methods are not suitable for resource constrained equipment, such as wireless sensor network (WSN) nodes, which have limited memory space and computation capability [5][6][7].…”