Flexible Quantization of Audio and Speech based on the Autoregressive Model

Ozerov, Alexey; Kleijn, W. Bastiaan

doi:10.1109/acssc.2007.4487270

Cited by 8 publications

(15 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…3.1) that we have already introduced in the partial case of AR model in [6]. Then, we introduce the new model estimation criterion (Sec.…”

Section: Rd Relations and New Estimation Criterionmentioning

confidence: 99%

“…Thus, we have to deal with a sequence of Gaussian models θ = {θ n } N n=1 , called hereafter model, that must be transmitted as well to the decoder as a side information 1 . Note that such a formulation covers both the GMMbased quantization [2,3], where a mixture index is transmitted as side information, and the AR model-based quantization [6], where quantized model parameters are transmitted.…”

Section: Rd Relations and New Estimation Criterionmentioning

confidence: 99%

“…The first scheme is a flexible coding scheme we recently reported in [6]. This scheme is based on the AR model and the KLT (the corresponding optimal adaptive transform).…”

Section: Ar Model Based Scheme With Kltmentioning

confidence: 99%

“…Using two coding schemes, our experiments confirm that the CR-MDL criterion outperforms the ML criterion. We analyze the case of a single Gaussian model varying in time, covering both the GMM-based quantization [2, 3] and the AR model-based quantization [6]. However, the framework is general, and can be extended to include other distributions, such as the generalized Gaussian distribution (GGD) , for which the ML criterion is also not optimal in the CR case.…”

Section: Introductionmentioning

confidence: 99%

“…For the case of a Gaussian mixture model (GMM) of the source, Subramaniam and Rao [2] introduced a practical CR quantization scheme, while [3] and [4] proposed practical solutions for the CE case. All these methods are based on high-rate (HR) theory approximations [5], but, as experiments show, they give satisfactory results for low rates as well [6]. In audio and speech coding, the techniques were successfully applied to the quantization of the line spectral frequencies (LSF) based on a GMM [2,4], to direct quantization of the signal based on a GMM [7], and to signal quantization using an adaptive autoregressive (AR) model [8,6].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Optimal parameter estimation for model-based quantization

Ozerov

Kleijn

2009

2009 IEEE International Conference on Acoustics, Speech and Signal Processing

Self Cite

View full text Add to dashboard Cite

We address optimal model estimation for model-based vector quantization for both the constrained resolution (CR) and constrained entropy (CE) cases. To this purpose we derive under high-rate (HR) theory assumptions the rate-distortion (RD) relations for these two quantization scenarios assuming a Gaussian model. Based on the RD relations we show that the maximum likelihood (ML) criterion leads to optimal performance for CE quantization, but not for CR quantization. We introduce a new model estimation criterion for CR quantization that is optimal (under HR theory assumptions) in terms of the RD relation. Our experiments confirm that the proposed criterion for model identification outperforms the ML criterion for a range of conditions.

show abstract

“…3.1) that we have already introduced in the partial case of AR model in [6]. Then, we introduce the new model estimation criterion (Sec.…”

Section: Rd Relations and New Estimation Criterionmentioning

confidence: 99%

Section: Rd Relations and New Estimation Criterionmentioning

confidence: 99%

“…The first scheme is a flexible coding scheme we recently reported in [6]. This scheme is based on the AR model and the KLT (the corresponding optimal adaptive transform).…”

Section: Ar Model Based Scheme With Kltmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Optimal parameter estimation for model-based quantization

Ozerov

Kleijn

2009

2009 IEEE International Conference on Acoustics, Speech and Signal Processing

Self Cite

View full text Add to dashboard Cite

show abstract

Predictive audio coding using rate-distortion-optimal pre- and post-filtering

Moussa

Kleijn

2011

2011 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA)

View full text Add to dashboard Cite

Flexcode - flexible audio coding

Klejsa

Kleijn

2010

2010 IEEE International Conference on Acoustics, Speech and Signal Processing

Self Cite

View full text Add to dashboard Cite

Modern networks are highly variable and, as a result, source coders are commonly used under conditions that they were not designed for. We address this problem with a source-coding philosophy that aims at the instantaneous re-optimization of a source coder to match a wide range of constraints on rate or quality and a wide range of packet-loss rates. We present a number of technologies that can be reconfigured by solving analytic relations that use the current conditions and a statistical description of the source as input. The technologies include distribution-preserving quantizers, flexible multiple-description quantizers, and a rate distribution scheme. Based on the generic technologies, we created a complete audio coder. Formal listening tests show that the resulting audio coding scheme with full flexibility provides a quality that is on-par with the best standardized codecs for any particular rate.Index Terms-audio coding, source modeling, quantization INTRODUCTIONThe telecommunication infrastructure is increasingly heterogeneous and time-variant. This paper describes a coding philosophy that aims to facilitate coder redesign in real-time. This means that coders that are based on this philosophy are always matched to the network and to quality constraints. The resulting coders are always optimal for the rate constraint and packet-loss conditions at hand. They can replace a range of existing standardized coders that have been designed and optimized for particular operating conditions. Typically, conceptually different approaches are used at different ranges of bit-rates. The best coders aiming at very low bit rates are parametric. Audio coders that are based on the CELP architecture (AMR-WB, G.729.1) provide the state-of-the-art performance at low to intermediate bit rates. State-of-the-art coders that operate at higher bit rates typically are transform-based (G.722.1, AAC) or use a hybrid CELP-transform architecture (AMR-WB+).In contrast to the state-of-the-art solutions, our coding approach is consistent. We propose a coding philosophy that relies on signal modeling and a well-defined distortion measure. This leads to technologies that can be optimized in real-time, and together these technologies yield an adaptive audio coder. The optimal configuration of the coder is found by solving sets of closed-form analytical expressions. In contrast to existing audio coders, our approach is optimal over a continuum of bit-rates.The paper is organized as follows. We first explain the philosophy of scalable coding in Section 2. We present a set of technologies in Section 3. In Section 4, we describe an exemplary implementation of the proposed coding scheme and Section 5 provides both objective performance results and the outcome of the formal listening test of a resulting coder. Conclusions are provided in Section 6.

show abstract

Flexible Quantization of Audio and Speech based on the Autoregressive Model

Cited by 8 publications

References 10 publications

Optimal parameter estimation for model-based quantization

Optimal parameter estimation for model-based quantization

Predictive audio coding using rate-distortion-optimal pre- and post-filtering

Flexcode - flexible audio coding

Contact Info

Product

Resources

About