Real-time implementations of sparse linear prediction for speech processing

Jensen, Tobias Lindstrøm; Giacobello, Daniele; Christensen, Mads Græsbøll; Jensen, Søren Holdt; Moonen, Marc

doi:10.1109/icassp.2013.6639260

Cited by 12 publications

(16 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[39, §11.8.2]. However, the reweighting destroys the Toeplitz structure and we need to resort to more general linear solvers like Cholesky factorization with time-complexity O(N 3 ), as in [30].…”

Section: An Admm Algorithm For the Slp Problemmentioning

confidence: 99%

“…This paper deals with understanding the trade-offs occurring in choosing a proper and fixed number of iterations for the ADMM algorithm and extends the analysis and algorithms presented in [30,38]. We note that we will apply an ADMM algorithm in its straightforward form but several variants and extensions may be useful for solving the sparse linear prediction problem efficiently and may be considered for further investigations.…”

Section: Introductionmentioning

confidence: 99%

“…Such diagonal matrices often destroys the possibility of faster direct methods for solving a linear system (see, e.g., the discussion in [28]). Specifically, IP methods for the SLP problem (5) would have cubic per iteration time complexity [29,30] and quadratic space complexity. In particular, the introduction of the diagonal scaling matrix for the SLP problem prohibits the exploration of Toeplitz structure since such matrices do not have a low displacement rank.…”

Section: Introductionmentioning

confidence: 99%

“…For these implementations, the GS approach was a faster choice than Levinson for k > 1. For comparison, a hand-tailored IP method[30] gavet k ≈ 1206 · 10 −6 + 2033 · 10 −6 k [s], C 2 = 0.999 . (29)As argued previously, one IP method iteration is much more expensive.…”

mentioning

confidence: 99%

See 3 more Smart Citations

Computational analysis of a fast algorithm for high-order sparse linear prediction

Jensen

Giacobello²,

Waterschoot

et al. 2016

2016 24th European Signal Processing Conference (EUSIPCO)

Self Cite

View full text Add to dashboard Cite

Using a sparsity promoting convex penalty function on high-order linear prediction coefficients and residuals has shown to result in improved modeling of speech and other signals as this addresses the inherent limitations of standard linear prediction methods. However, this new formulation is computationally more demanding which may limit its use, in particular for embedded signal processing. This paper analyzes the algorithmic and computational aspects of the matrix structures associated with an alternating direction method of multipliers algorithm for solving the convex high-order sparse linear prediction problem. The paper also analyzes the inherent trade-off between accuracy and the objective measure of prediction gain and shows that a few iterations are sufficient to achieve similar results as computationally more expensive interior-point methods.

show abstract

Section: An Admm Algorithm For the Slp Problemmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

Computational analysis of a fast algorithm for high-order sparse linear prediction

Jensen

Giacobello²,

Waterschoot

et al. 2016

2016 24th European Signal Processing Conference (EUSIPCO)

Self Cite

View full text Add to dashboard Cite

show abstract

“…This optimisation problem can be solved by various methods (Eg: interior point methods, iterative re−weighted 1 norm minimisation algorithm etc). Here primal-dual interior-point method is used to solve this optimisation problem [12], because it is very efficient for solving convex optimization problems in real-time applications. Database is created using the sparse LP coefficient matrix obtained from the training speech samples.…”

Section: A Feature Extractionmentioning

confidence: 99%

Sparse linear prediction coefficients for isolated speech recognition

Baburaj

George

2015

2015 International Conference on Control Communication &Amp; Computing India (ICCC)

View full text Add to dashboard Cite

Isolated speech recognition system is an important step in many applications such as automated banking system, catalogue dialing, automated data entry, robotics etc. Selection of feature and classifier in the speech recognition system is based on the complexity and recognition accuracy. Mel-frequency cepstral coefficients (MFCCs), line spectral frequencies (LSF), short time energy (STE) and linear prediction coefficients (LPC) are the features used in the existing speech recognition systems. In this paper, a sparse feature, obtained from the optimization of linear prediction coefficients (LPC) with a sparsity constraint is used for the classification. These sparse linear prediction coefficients (sparse LPC) offer a more effective way of representing the voiced speech. Artificial neural network (ANN) is used for the classification purpose. Experimental results show that the proposed method is noise robust and its performance exceeds LPC and MFCC feature based speech recognition systems.

show abstract

Fast algorithms for high-order sparse linear prediction with applications to speech processing

Jensen

Giacobello

Waterschoot

et al. 2016

Speech Communication

Self Cite

View full text Add to dashboard Cite

In speech processing applications, imposing sparsity constraints on high-order linear prediction coefficients and prediction residuals has proven successful in overcoming some of the limitation of conventional linear predictive modeling. However, this modeling scheme, named sparse linear prediction, is generally formulated as a linear programming problem that comes at the expenses of a much higher computational burden compared to the conventional approach. In this paper, we propose to solve the optimization problem by combining splitting methods with two approaches: the Douglas-Rachford method and the alternating direction method of multipliers. These methods allow to obtain solutions with a higher computational efficiency, orders of magnitude faster than with general purpose software based on interior-point methods. Furthermore, computational savings are achieved by solving the sparse linear prediction problem with lower accuracy than in previous work. In the experimental analysis, we clearly show that a solution with lower accuracy can achieve approximately the same performance as a high accuracy solution both objectively, in terms of prediction gain, as well as with perceptually relevant measures, when evaluated in a speech reconstruction application.

show abstract

Real-time implementations of sparse linear prediction for speech processing

Cited by 12 publications

References 19 publications

Computational analysis of a fast algorithm for high-order sparse linear prediction

Computational analysis of a fast algorithm for high-order sparse linear prediction

Sparse linear prediction coefficients for isolated speech recognition

Fast algorithms for high-order sparse linear prediction with applications to speech processing

Contact Info

Product

Resources

About