scispace - formally typeset
Search or ask a question
Topic

Filter bank

About: Filter bank is a research topic. Over the lifetime, 10465 publications have been published within this topic receiving 189911 citations. The topic is also known as: filterbank.


Papers
More filters
Proceedings ArticleDOI
01 Dec 2016
TL;DR: Experimental results on several benchmark datasets show that the superior performance of the proposed neural network based clothing style analysis neural network (CSANN) method is superior to existing methods.
Abstract: A neural network based clothing style analysis method is proposed via deep filter bank in this paper. Clothing styles are complicated and high-level concept. We propose to construct the deep filter bank by combining Convolution Neural Network(CNN) with Fisher Vector(FV). Then, the extracted features from the body part are used to train the Part-CNN (p-CNN) model. Multiple p-CNNs are integrated along with the whole body model to train the Clothing Style Analysis Neural Network (CSANN). Experimental results on several benchmark datasets show that the superior performance of the proposed method.
Book ChapterDOI
18 Sep 2006
TL;DR: In this paper, a method for designing 2D non-separable filter banks with the LP and near perfect reconstruction (near-PR) properties was proposed by combining the unimodular transformation and a separable |M|-channel LP filter bank, the design problem is simplified to that of two one-dimensional (1D) LP filter banks.
Abstract: Two dimensional (2D) nonseparable filter banks with linear phase (LP) are desired for image sub-band coding and compression. In this paper, we propose a method for designing 2D nonseparable filter banks with the LP and near perfect reconstruction (near-PR) properties. By combining the unimodular transformation and a separable |M|-channel LP filter bank, the design problem is simplified to that of two one-dimensional (1D) LP filter banks. For 1D case, a novel method by employing partial cosine modulation is used to design near-PR filter banks with LP analysis and synthesis filters. For 2D case, we cascade two 1D near-PR LP filter banks in the form of tree structure to design a separable LP filter bank. With the unimodular transformation, a nonseparable LP filter bank is obtained. In addition, the filter bank achieves the near-PR property without sophisticated nonlinear optimization procedures. Design example shows the efficiency and simplicity of the proposed method.
Patent
28 Oct 2004
TL;DR: In this article, the authors provided a method for encoding an audio signal, including a first audio channel and a second audio channel encoding of the plurality for a first plurality of subband signals and the second audio channels for the first audio channels.
Abstract: It is provided to encoding an audio signal, the audio signal, including a first audio channel and a second audio channel encoding of the plurality for a first plurality of subband signals and the second audio channel for the first audio channel the method comprising the second sub first audio channel and a second audio channel of each of the composite modular filter bank to provide the band signals sub-band filtering; Of down-sampled plurality of first sub-band signals and down-sampling the second plurality to provide subband signals comprising: downsampling each of the subband signals; Adding subband filtering at least one of the downsampled subband signals of the additional filter bank for providing subband signals, the plurality of sub; The down-sampled sub-band signals without further subband filtering and a sub-step for deriving spatial parameters from the sub-band signal; Of down-sampled plurality of first sub-band signals and down-sampling a second plurality comprises the steps of inducing a single channel audio signal comprising derived subband signals derived from the subband signals Further, the decoding is provided, the encoded audio signal is set by including an encoded single channel audio signal and spatial parameters are decoded by a single channel audio channel encoded to obtain a plurality of downsampled subband signals, and a plurality of sub-portion of at least one of down-sampled sub-band signals in a filter bank of the addition to provide a sub-band signal sub-band filter, and deriving two audio channels from the spatial parameters, the sub-subband and the signals down-sampled subband signals are no sub-band filtering of the addition
Patent
07 Jan 2016
TL;DR: In this article, a transmission apparatus is adapted to generate at least one resource block from a time-frequency resource grid corresponding to the payload signal, where each resource block is a specific spectral section for a specific time slot, and comprises one spectral mid-section and one spectral border section.
Abstract: A transmission apparatus (30) serves the purpose of generating a filter bank multicarrier FBMC signal (32) from a payload signal (31). The transmission apparatus (30) comprises an offset quadrature amplitude modulator (41) and a quadrature amplitude modulator (42). The transmission apparatus (30) is adapted to generate at least one resource block from a time-frequency resource grid corresponding to the payload signal (31), wherein each resource block is a specific spectral section for a specific time slot, and comprises one spectral mid-section and at least one spectral border-section being located at a spectral boarder of each resource block. The offset quadrature amplitude modulator (41) is adapted to generate the spectral mid-sections of each resource block by performing an offset quadrature amplitude modulation to a first part of the payload signal (31). The quadrature amplitude modulator (42) is adapted to generate the at least one spectral border section of each resource block by performing a quadrature amplitude modulation to a second part of the payload signal (31).

Network Information
Related Topics (5)
Feature extraction
111.8K papers, 2.1M citations
88% related
Robustness (computer science)
94.7K papers, 1.6M citations
85% related
Image segmentation
79.6K papers, 1.8M citations
85% related
Wireless
133.4K papers, 1.9M citations
83% related
Convolutional neural network
74.7K papers, 2M citations
83% related
Performance
Metrics
No. of papers in the topic in previous years
YearPapers
202361
2022188
2021174
2020239
2019289
2018309