Vector Quantization of Harmonic Magnitudes in Speech Coding Applications—A Survey and New Technique

Abstract

A harmonic coder extracts the harmonic components of a signal and represents them efficiently using a few parameters. The principles of harmonic coding have become quite successful and several standardized speech and audio coders are based on it. One of the key issues in harmonic coder design is in the quantization of harmonic magnitudes, where many propositions have appeared in the literature. The objective of this paper is to provide a survey of the various techniques that have appeared in the literature for vector quantization of harmonic magnitudes, with emphasis on those adopted by the major speech coding standards; these include constant magnitude approximation, partial quantization, dimension conversion, and variable-dimension vector quantization (VDVQ). In addition, a refined VDVQ technique is proposed where experimental data are provided to demonstrate its effectiveness.

A new perceptually weighted distance measure for vector quantization of the STFT amplitudes in the speech application

“…Besides; in order to determine the STFT amplitudes, the hamming window and 128-point FFT are utilized. After designing the codebook, the spectral distortion (SD) criterion [4,8] is utilized for the evaluation of the VQ process. In the evaluation phase, 20 (4 × 5) different sentences of same speakers are used.…”

“…The optimization algorithm varies α m and α h around the above initial values and satisfies the constraint of 0 < α h < α m < α l = 1. In order to determine the optimum values of α m and α h , the optimization algorithm utilizes the spectral distortion (SD) criterion [4,8]. In other words, the optimum values of α m and α h lead to the least spectral distortion in our simulations.…”

“…In [3], making use of the VQ, the LPC parameters are quantized and the quantized parameters are used for the speech enhancement. In [4], the VQ is applied to the harmonic amplitude of the speech. Moreover, in [5], LSF parameters are quantized by a perceptually weighted distance function.…”

Multi-pitch estimation of audio recordings using a codebook-based approach

“…The results of the estimator could be used to initialize the EM algorithm [11,10], and the method could be used in automatic music transcription, enhancement, and separation systems. Future work includes investigating the choice of the number of harmonic amplitudes to include in the codebook, e.g., by using a technique such as variable-dimension vector quantization (VDVQ) [23]. Furthermore it should be investigated whether the amplitudes can be modeled statistically instead of using a codebook approach, which involves training, e.g., by using linear prediction [24].…”

A simulation study of harmonics regeneration in noise reduction for electric and acoustic stimulation

“…Since voiced frames have much higher local SNRs compared with unvoiced frames, the estimation of voiced cues is generally much more robust. Many robust F0 detection algorithms exist and a harmonic regeneration operator can be based on adaptive comb filtering ͑Ne-horai and Porat, 1986͒, nonlinear functions ͑Plapous et al, 2006͒ and codebook approach ͑because harmonics amplitudes are highly correlated; see more details in Chu, 2004;Zavarehei et al, 2007, Fig. 2͒.…”