Descend-Delta-Mean Algorithm for Feature Extraction of Isolated THAI Digit Speech

Nosan, A.; Sitjongsataporn, Suchada

doi:10.1109/ieecon45304.2019.8938903

Cited by 2 publications

(6 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…N is the amount of input data and the size-dependent on framing. The iteration of FFT is called nFFT, and then is the index sequence of /2 [3,13]. As considered the result of Eq.…”

Section: Short Time Fourier Transformmentioning

confidence: 99%

“…The ASDM-MFCC algorithm is to find the average absolute value of a delta of a descendant as follows. The ASDM-MFCC algorithm was modified and developed from [3] and [13]. The modified algorithm is shown an acoustic property that can withstand noise at low SNR without affecting system performance and to increase robustness against the different types of environmental noises.…”

Section: Mel Frequency Cepstral Coefficients (Mfcc)mentioning

confidence: 99%

“…To increase the accuracy rate of speech and speaker recognition, the characteristic speech signal represented more resistance to noise than another algorithm. The descend-delta-mean algorithm (DDM) has been proposed [3] to against the interference and noise. The End-to-End noisy speech recognition using Fourier and Hilbert spectrum features [4] has been improved the noise robustness by adding components to the recognition system.…”

Section: Introductionmentioning

confidence: 99%

“…That introduces the improving the feature extraction method of the MFCC algorithm for noise robustness when there is interference in the speech input signal. It was modified and developed from the DDM algorithm [3] as the descend-delta-mean and mel-frequency cepstral coefficients (DDM-MFCC). The purpose is to require the noise robustness of the feature vector extracted from the speech signal with the characteristic coefficients of the proposed algorithm.…”

Section: Introductionmentioning

confidence: 99%

“…To overcome these problems above, in this paper to achieve satisfactory performance in speech recognition, a new method to obtain speech feature extraction for noise robustness is proposed. It will be modified and developed from the MFCC [3] and the DDM-MFCC [13] algorithm. This has been modified to provide the properties that can withstand noise at low SNR without affecting system performance.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Enhanced Feature Extraction Based on Absolute Sort Delta Mean Algorithm and MFCC for Noise Robustness Speech Recognition

Nosan¹,

Sitjongsataporn²

2021

IJIES

Self Cite

View full text Add to dashboard Cite

In this paper, a proposed absolute sort delta mean (ASDM) method obtaining the speech feature extraction for noise robustness is developed from mel-frequency cepstral coefficients (MFCC) named ASDM-MFCC, in order to increase robustness against the different types of environmental noises. This method is used to suppress the noise effects by finding a rearranging average of power spectrum magnitude combined with triangular bandpass filtering. Firstly, the spectral power magnitudes are sorted in each frequency band of the speech signal. Secondly, the absolutedelta values are arranged and then a mean value is determined in the last step. The purpose of proposed ASDM-MFCC algorithm is to require the noise robustness of the feature vector extracted from the speech signal with the characteristic coefficients. The NOIZEUS noisy speech corpus dataset is used to evaluate the performance of proposed ASDM-MFCC algorithm by Euclidean distance method with the low computation complexity. Experimental results show that the proposed method can provide significantly the improvement in terms of accuracy at low signal to noise ratio (SNR). In the case of car and station at SNR=5dB, the proposed approach can outperform in comparison with the conventional MFCC and gammatone frequency cepstral coefficient (GFCC) by 80% and 76.67%, respectively. Obviously, some experimental results of the proposed ASDM-MFCC algorithm are more robust than the traditional one.

show abstract

“…N is the amount of input data and the size-dependent on framing. The iteration of FFT is called nFFT, and then is the index sequence of /2 [3,13]. As considered the result of Eq.…”

Section: Short Time Fourier Transformmentioning

confidence: 99%

Section: Mel Frequency Cepstral Coefficients (Mfcc)mentioning

confidence: 99%