Compressive sensing based scalable speech coder with dynamic selection of basis and vector quantization

Sankar, Meena; Sathidevi, P. S.

doi:10.1109/wispnet.2017.8299923

Cited by 1 publication

(6 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We conducted experiments using various combination of features and selected ZCR and energy of frames for the categorization as proposed in our previous work due to its performance [12] .…”

Section: Determination Of Various Categories For Speechmentioning

confidence: 99%

“…Hence, the reconstruction of speech samples using the hybrid basis of wavelet db8 and DCT gives the minimum average basis count and has been selected as the basis for voiced speech components. Unvoiced Our previous work on dynamic basis selection [12] , [16] used LPC basis for mapping unvoiced sounds. In [25] , the author used Daubechies wavelet of order 3 due to the high-frequency nature for analysis and representation of unvoiced sounds.…”

Section: Identification Of Optimum Basis For Each Categorymentioning

confidence: 99%

“…In speech processing, CS is used for coding, enhancement and encryption [11] , [12] . Daniels et al [8] proposed a scalable Algebraic Code Excited Linear Prediction (ACELP) using the compressed framework which offers scalability by varying the dimensions of measurements.…”

Section: Introductionmentioning

confidence: 99%

“…In [13] , a hybrid dictionary of Linear predictive Coding (LPC) and Discrete Cosine Transform (DCT) is proposed for the representation of speech signal to account for the time-varying nature. In our previous work [12] , we did a dynamic selection of basis as per the statistical properties of the signal. In [14] , the CS concept is used for both speech coding and enhancement by exploiting its sparsity in the frequency domain.…”

Section: Introductionmentioning

confidence: 99%

“…The quantization error of measurements has a great impact on the reconstruction accuracy due to its significance in sparse recovery. In our previous works [12] , [16] , the probability distribution of speech is exploited for quantization of measurements but it requires both the mean and variance of each measurement vector to be transmitted along with indices. In [14] , the quantization of measurements is done using Analysis-by-synthesis (AbS) approach which requires the reconstruction of signal at transmitter that increases its complexity.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

A scalable speech coding scheme using compressive sensing and orthogonal mapping based vector quantization

Sankar

Sathidevi

2019

Heliyon

View full text Add to dashboard Cite

A novel scalable speech coding scheme based on Compressive Sensing (CS), which can operate at bit rates from 3.275 to 7.275 kbps is designed and implemented in this paper. The CS based speech coding offers the benefit of combined compression and encryption with inherent de-noising and bit rate scalability. The non-stationary nature of speech signal causes the recovery process from CS measurements very complex due to the variation in sparsifying bases. In this work, the complexity of the recovery process is reduced by assigning a suitable basis to each frame of the speech signal based on its statistical properties. As the quality of the reconstructed speech depends on the sensing matrix used at the transmitter, a variant of Binary Permuted Block Diagonal (BPBD) matrix is also proposed here which offers a better performance than that of the commonly used Gaussian random matrix. To improve the coding efficiency, formant filter coefficients are quantized using the conventional Vector Quantization (VQ) and an orthogonal mapping based VQ is developed for the quantization of CS measurements. The proposed coding scheme offers the listening quality for reconstructed speech similar to that of Adaptive Multi rate-Narrowband (AMR-NB) codec at 6.7 kbps and Enhanced Voice Services (EVS) at 7.2 kbps. A separate de-noising block is not required in the proposed coding scheme due to the inherent de-noising property of CS. Scalability in bit rate is achieved in the proposed method by varying the number of random measurements and the number of levels for orthogonal mapping in the VQ stage of measurements.

show abstract

“…We conducted experiments using various combination of features and selected ZCR and energy of frames for the categorization as proposed in our previous work due to its performance [12] .…”

Section: Determination Of Various Categories For Speechmentioning

confidence: 99%