Enhanced Signal Classifier for Speech/Audio Coding by Binary Decision Tree
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Various researches are being conducted on an integrated speech/audio coder capable of handling both speech and audio signals. However, since the speech and the audio signals have different characteristics, it is difficult to process them with a single coder. Therefore, instead of coding the speech and the audio with only one coder, a method of coding the speech with a speech coder and the audio with an audio coder was adopted, and an integrated speech/audio coder combining the speech and the audio coders was developed. That is, after signal classification is performed on the input signal, the speech is coded by the speech coder and the audio is coded by the audio coder. G.718 is a representative speech/audio coder that uses a signal classifier using modified discrete cosine transform (MDCT) energy ratio to classify input signals into the speech and the audio and then codes the input signals. However, since the signal classifier of G.718 performs a simple signal classification using a single parameter, the performance of the speech/audio signal classifier is not good, resulting in degradation of the overall performance of the speech/audio coder. Therefore, in this paper, a speech/audio signals classifier for the frame-by frame signal coding by a new decision rule is proposed. In the proposed classification, various time and frequency domain features are extracted. To design an optimized speech/audio classifier, the classification and regression tree (CART) algorithm is utilized and a binary decision tree is constructed. The performance of the proposed classifier with the various features is compared with that of the G.718 with the ratio between the MDCT power ratio of the input and the code-excited linear prediction (CELP) coder output signal. Experimental results show that the classification rates of the proposed algorithm are higher than those of the classifiers in G.718.
Abstract
Various researches are being conducted on an integrated speech/audio coder capable of handling both speech and audio signals. However, since the speech and the audio signals have different characteristics, it is difficult to process them with a single coder. Therefore, instead of coding the speech and the audio with only one coder, a method of coding the speech with a speech coder and the audio with an audio coder was adopted, and an integrated speech/audio coder combining the speech and the audio coders was developed. That is, after signal classification is performed on the input signal, the speech is coded by the speech coder and the audio is coded by the audio coder. G.718 is a representative speech/audio coder that uses a signal classifier using modified discrete cosine transform (MDCT) energy ratio to classify input signals into the speech and the audio and then codes the input signals. However, since the signal classifier of G.718 performs a simple signal classification using a single parameter, the performance of the speech/audio signal classifier is not good, resulting in degradation of the overall performance of the speech/audio coder. Therefore, in this paper, a speech/audio signals classifier for the frame-by frame signal coding by a new decision rule is proposed. In the proposed classification, various time and frequency domain features are extracted. To design an optimized speech/audio classifier, the classification and regression tree (CART) algorithm is utilized and a binary decision tree is constructed. The performance of the proposed classifier with the various features is compared with that of the G.718 with the ratio between the MDCT power ratio of the input and the code-excited linear prediction (CELP) coder output signal. Experimental results show that the classification rates of the proposed algorithm are higher than those of the classifiers in G.718.
참고문헌 (16)
- M. Jelinek, "ITU-T G.EV-VBR baseline codec", IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp.4749~4752 (2008)
- T. Vaillancourt, "Advances in low bitrate time-frequency coding", IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (2015)
- "ITU-T Recommendations on G.718, Frame error robust narrowband and wideband embedded variable bit-rate coding of speech and audio from 8–32 kb", s, 2008
- M. Dietz, "Overview of the EVS codec architecture", 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (2015)
- W. B. Kleijin, "Generative speech coding with predictive variance regularization", 2021IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (2021), IEEE
- M. Jelinek, "G. 718 : A new embedded speech and audio coding standard with high resilience to error-prone transmission channels", IEEE Communications Magazine, 46(10), pp.117~123 (2009)
- J. Wang, "A temporal-spectral generative adversarial network based end-to-end packet loss concealment for wideband speech transmission", The Journal of the Acoustical Society of America, 150(4), pp.2577~2588 (2021)
- G. Markovic, "Transform-based Coding Methods for Speech and other Audio Signals", Doctoral, pp.2022
- S. Furui, "Advances in speech signal processing" (1991), Marcel Dekker Inc
- G. Tzanetakis, "Musical genre classification of audio signals", IEEE Transactions on Speech and Audio Processing, 10(5), pp.293~302 (2002)
- Z. Fu, "A survey of audio-based music classification and annotation", IEEE transactions on multimedia, 13(2), pp.303~319 (2010)
- G. Sharma, "Trends in audio signal feature extraction methods", Applied Acoustics, 158 (2020)
- D. Cournapeau, "Evaluation of real-time activity detection based on high order statistics", pp.2945~2948 (2007), Interspeech
- J. F. Elder, "A comparison of leading data mining tools", 4th International Conference on Knowledge Discovery and Data Mining, pp.1~31 (1998)
- Y. Kim, "A GMM-based target classification scheme for a node in wireless sensor networks", IEICE Transactions on Communications, E91-B(11) (2008)
- "ITU-T Recommendations on BS. 1534, Method for the subjective assessment of intermediate quality levels of coding systems", ~ (2003)