Practical Compression Methods for Quantized Compressed Sensing

Leinonen, Markus; Codreanu, Marian; Juntti, Markku

doi:10.1109/infcomw.2019.8845224

Cited by 5 publications

(15 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…• A compress-and-estimate (CE) QCS method [20], [21] where 1) the encoder quantizes measurements y in (1) oblivious to x, and 2) the decoder estimates x from quantized measurementsỹ ∈ R M through a quadratically constrained polynomial-complexity basis pursuit (BP) problem 20 min.…”

Section: ) Baseline Qcs Methodsmentioning

confidence: 99%

“…• An estimate-and-compress (EC) QCS method, EC-VQ [20], [21], where 1) the encoder forms an MMSE estimate of x from y which is an exponentially complex task [57], and 2) quantizes the resulting estimate with a Lloydoptimized VQ. The EC strategy is known to be the optimal compression strategy for remote source coding [14], [30], while suffering from its high complexity.…”

Section: ) Baseline Qcs Methodsmentioning

confidence: 99%

“…To assess the computational complexity and latency for communicating a measurement vector y, Table I compares the algorithm running time 21 in the online phase for three different signal setups. The decoding and total running times of the proposed DeepVQCS scheme are around 35 − 60 and 25 − 40 times lower than those of the polynomial-complexity CE-USQ-L1 method, respectively.…”

Section: ) Algorithm Running Timementioning

confidence: 99%

“…At the cost of increased complexity, enhanced rate-distortion performance is achieved by vector quantization (VQ) [18]- [20]. For empirical performance comparison of various QCS methods, see e.g., [21], [2,Sect. 7.6].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Low-Complexity Vector Quantized Compressed Sensing via Deep Neural Networks

Leinonen¹,

Codreanu²

2020

IEEE Open J. Commun. Soc.

Self Cite

View full text Add to dashboard Cite

Sparse signals, encountered in many wireless and signal acquisition applications, can be acquired via compressed sensing (CS) to reduce computations and transmissions, crucial for resource-limited devices, e.g., wireless sensors. Since the information signals are often continuous-valued, digital communication of compressive measurements requires quantization. In such a quantized compressed sensing (QCS) context, we address remote acquisition of a sparse source through vector quantized noisy compressive measurements. We propose a deep encoder-decoder architecture, consisting of an encoder deep neural network (DNN), a quantizer, and a decoder DNN, that realizes low-complexity vector quantization aiming at minimizing the mean-square error of the signal reconstruction for a given quantization rate. We devise a supervised learning method using stochastic gradient descent and backpropagation to train the system blocks. Strategies to overcome the vanishing gradient problem are proposed. Simulation results show that the proposed non-iterative DNN-based QCS method achieves higher rate-distortion performance with lower algorithm complexity as compared to standard QCS methods, conducive to delay-sensitive applications with large-scale signals.

show abstract

Section: ) Baseline Qcs Methodsmentioning

confidence: 99%

Section: ) Baseline Qcs Methodsmentioning

confidence: 99%

Section: ) Algorithm Running Timementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Low-Complexity Vector Quantized Compressed Sensing via Deep Neural Networks

Leinonen¹,

Codreanu²

2020

IEEE Open J. Commun. Soc.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Sparsity in the sensed data allows to reduce the computations of simple sensor devices, and most importantly, to significantly reduce sensors' energy consumption for communicating the data to a fusion center [175,176,178]. One particular direction of great interest is so-called quantized CS [129,273,333,151,180,184,182,181] -a lossy compression setup where the CS measurements are converted into finite-rate bit sequences, and the aim is to design efficient quantization-aware CS reconstruction algorithms and analyze their ratedistortion performance. Application requirements may demand as low quantization rate as one bit per measurement sample, referred as 1-bit CS [152].…”

Section: Signal Sparsity In the Considered Applicationsmentioning

confidence: 99%

Compressed Sensing with Applications in Wireless Networks

Leinonen

Codreanu

Giannakis³

2019

FNT in Signal Processing

Self Cite

View full text Add to dashboard Cite

Sparsity is an attribute present in a myriad of natural signals and systems, occurring either inherently or after a suitable projection. Such signals with lots of zeros possess minimal degrees of freedom and are thus attractive from an implementation perspective in wireless networks. While sparsity has appeared for decades in various mathematical fields, the emergence of compressed sensing (CS)-the joint sampling and compression paradigm-in 2006 gave rise to plethora of novel communication designs that can efficiently exploit sparsity. In this monograph, we review several CS frameworks where sparsity is exploited to improve the quality of signal reconstruction/detection while reducing the use of radio and energy resources by decreasing, e.g., the sampling rate, transmission rate, and number of computations. The first part focuses on several advanced CS signal reconstruction techniques along with wireless applications. The second part deals with efficient data gathering and lossy compression techniques in wireless sensor networks. Finally, the third part addresses CS-driven designs for spectrum sensing and multiuser detection for cognitive and wireless communications.

show abstract

Quantized Compressed Sensing via Deep Neural Networks

Leinonen

Codreanu

2020

2020 2nd 6G Wireless Summit (6G SUMMIT)

Self Cite

View full text Add to dashboard Cite

Compressed sensing (CS) is an efficient technique to acquire sparse signals in many wireless applications to, e.g., reduce the amount of data and save low-power sensors' batteries. This paper addresses efficient acquisition of sparse sources through quantized noisy compressive measurements where the encoder and decoder are realized by deep neural networks (DNNs). We devise a DNN based quantized compressed sensing (QCS) method aiming at minimizing the mean-square error of the signal reconstruction. Once trained offline, the proposed method enjoys extremely fast and low complexity decoding in the online communication phase. Simulation results demonstrate the superior rate-distortion performance of the proposed method compared to a polynomial-complexity QCS reconstruction scheme.

show abstract

Practical Compression Methods for Quantized Compressed Sensing

Cited by 5 publications

References 35 publications

Low-Complexity Vector Quantized Compressed Sensing via Deep Neural Networks

Low-Complexity Vector Quantized Compressed Sensing via Deep Neural Networks

Compressed Sensing with Applications in Wireless Networks

Quantized Compressed Sensing via Deep Neural Networks

Contact Info

Product

Resources

About