Proceedings of ICASSP '94. IEEE International Conference on Acoustics, Speech and Signal Processing
The paper presents an improved collage theorem, valid for a class of signal mappings called Afine... more The paper presents an improved collage theorem, valid for a class of signal mappings called Afine Blockwise Aueraging (ABA) Mappings. The .4BA structure is exploited to form a bound depending on norms of collage error signals at several resolutions. Compared to previously published collage theorems, the new theorem provides a much tighter bound on the maximum distance between the original signal (real world image) and the decoded attractor, given the distance between the original and the "collage" optimized by the encoder. This is achieved without contractivity constraints on the .4BA mapping parameters. Finally, an encoding method in which one attempts to minimize the upper bound directly is suggested.
2007 IEEE International Conference on Image Processing, 2007
Show-Through is a common occurrence when scanning duplex printed documents. The back-side printin... more Show-Through is a common occurrence when scanning duplex printed documents. The back-side printing shows through the paper, contaminating the front side image. Previous work modeled the problem as a non-linear convolutive mixture of images and offered solutions based on decorrelation. In this work we propose a cleaning process based on a Blind Source Separation approach. We define a cost function incorporating the non-linear mixing model in a mean-squared error term, along with a regularization term based on Total-Variation. We propose a location dependent regularization tradeoff, preserving image edges while removing show-through edges. The images and mixing parameters are estimated using an alternating minimization process, with each stage using only convex optimization methods. The resulting images exhibit significantly lower show-through, both visibly and in objective measures.
We present a generalization of the concept of Convex Sets, based on the Morphological Closing ope... more We present a generalization of the concept of Convex Sets, based on the Morphological Closing operation, and study some of its properties. We also de ne Extreme Points of such Generalized Convex Sets, which generalize the notion of Extreme Points of Convex Sets. Moreover, we apply the above notions to skeleton redundancy removal, and present an algorithm for obtaining an Error-Free Skeleton representation with reduced amount of redundant points, using morphological operations only.
2017 IEEE International Conference on Computer Vision (ICCV), 2017
Local-to global point cloud registration is a challenging task due to the substantial differences... more Local-to global point cloud registration is a challenging task due to the substantial differences between these two types of data, and the different techniques used to acquire them. Global clouds cover large-scale environments and are usually acquired aerially, e.g., 3D modeling of a city using Airborne Laser Scanning (ALS). In contrast, local clouds are often acquired from ground level and at a much smaller range, for example, using Terrestrial Laser Scanning (TLS). The differences are often manifested in point density distribution, occlusions nature, and measurement noise. As a result of these differences, existing point cloud registration approaches, such as keypoint-based registration, tend to fail. We improve upon a different approach, recently proposed, based on converting the global cloud into a viewpoint-based cloud dictionary. We propose a local-toglobal registration method where we replace the dictionary clouds with viewpoint descriptors, consisting of panoramic range-imag...
The Optimized Temporal Decomposition (OTD) technique for Line Spectral Frequencies (LSF) speech e... more The Optimized Temporal Decomposition (OTD) technique for Line Spectral Frequencies (LSF) speech envelope representation, under a MMSE criterion, has been shown to be promising for very low bit rate speech coding for storage and broadcast applications. In order to improve perceptual speech quality, a dynamically weighted OTD (DW-OTD) technique is introduced in this work. It extends the OTD by allowing temporally changing weights, so as to improve the perceived speech quality. Use of Gardner's weighted MSE with DWOTD is found to reduce the Log Spectral Distance (LSD) measure by 0.3 dB, as compared to OTD. The original OTD algorithm delay and complexity requirements make it inappropriate for real-time speech coding. In this paper we also introduce a modification of this technique, which is suboptimal but suitable for on-line speech coding purposes, with negligible degradation of performance (of only about 0.06 dB in LSD). With the proposed techniques we were able to encode speech s...
As 3D point cloud acquisition sensors become increasingly prevalent in urban environments (e.g., ... more As 3D point cloud acquisition sensors become increasingly prevalent in urban environments (e.g., LiDAR sensors for autonomous vehicles), the need arises to find efficient ways to align large amounts of such 3D data, often in real-time. In this work, we propose a novel method for 3D point cloud registration refinement in an urban environment (e.g., between Terrestrial LiDAR Scans - TLS - and Airborne LiDAR Scans - ALS), assuming an initial coarse registration is available. The proposed method is based on estimation of the direction of gravity, wall detection, projection of the point clouds on a perpendicular horizontal plane, and conversion into 2D edge-maps. Then, two methods are considered for alignment between the 2D edge-maps: a 2D variant of the well-known ICP (Iterative Closest Point) algorithm, and Edge-Map Phase-Correlation (EMPC). We demonstrate the usefulness of the proposed methods for registration in this challenging task, where the 2D variant of ICP achieves a meaningful...
We present a new tree-based framework for producing self-dual morphological operators. For any gi... more We present a new tree-based framework for producing self-dual morphological operators. For any given tree representation of images, one can associate a complete inf-semilattice (CISL) in the corresponding tree-representation domain, where the operators can then be derived. We also present a particular case of this general framework, involving a new tree representation, the ExtremaWatershed Tree (EWT). The operators obtained by using the EWT in the above framework behave like classical morphological operators, but in addition are self-dual. Some application examples are provided: pre-processing for OCR and dust & scratch removal algorithms, and image denoising.
Mobile devices are widely used around the world, frequently by people speaking local languages or... more Mobile devices are widely used around the world, frequently by people speaking local languages or dialects that are not well documented. For these languages, it might not be beneficial for commercial companies to develop Automatic Speech Recognition (ASR) systems, so users of these languages cannot utilize voice activation features (often using Keyword Spotting, KWS) of their devices. Standard KWS methods aim to statistically model the generation process of the speech signal, requiring hours of recorded and transcribed speech for training, and therefore are not adequate for limited-data scenarios. In this paper we propose a new KWS method, suitable for limited-data scenarios, which can be easily applied by developers. The proposed method uses a new histogram representation for words, obtained with respect to a pre-trained Gaussian Mixture Model (GMM). Sentences are represented by fixed-length global feature vectors, extracted from the response curves obtained by a word classifier. Word and sentence classifiers are trained using a discriminative approach, which is typically robust to training-set size. The dataset for training the GMM is easy to obtain, since no annotation is required. We compared the proposed system to a Hidden Markov Model (HMM) based system, trained using the same low data-resources conditions as ours, and to a state-of-the-art ASR system, trained using either the limited data scenario, or using many hours of recorded speech. In the limited data situation, our system performs better then both benchmarks in all experiments except for clean speech of children (CSLU dataset), where it performs as good as the HMM. Since the ASR benchmark performs poorly without enough training data, we also trained it without limiting the available data. In this case the ASR benchmark performs better when tested on speech of adults (TED-LIUM dataset of TED lectures) for all noise conditions, and our system performs better when tested on speech of children with low to moderate SNR values. The results demonstrate the advantages of the proposed system, and the conditions under which it performs better.
EURASIP Journal on Audio, Speech, and Music Processing, 2016
The goal of voice conversion is to modify a source speaker's speech to sound as if spoken by a ta... more The goal of voice conversion is to modify a source speaker's speech to sound as if spoken by a target speaker. Common conversion methods are based on Gaussian mixture modeling (GMM). They aim to statistically model the spectral structure of the source and target signals and require relatively large training sets (typically dozens of sentences) to avoid over-fitting. Moreover, they often lead to muffled synthesized output signals, due to excessive smoothing of the spectral envelopes. Mobile applications are characterized with low resources in terms of training data, memory footprint, and computational complexity. As technology advances, computational and memory requirements become less limiting; however, the amount of available training data still presents a great challenge, as a typical mobile user is willing to record himself saying just few sentences. In this paper, we propose the grid-based (GB) conversion method for such low resource environments, which is successfully trained using very few sentences (5-10). The GB approach is based on sequential Bayesian tracking, by which the conversion process is expressed as a sequential estimation problem of tracking the target spectrum based on the observed source spectrum. The converted Mel frequency cepstrum coefficient (MFCC) vectors are sequentially evaluated using a weighted sum of the target training vectors used as grid points. The training process includes simple computations of Euclidian distances between the training vectors and is easily performed even in cases of very small training sets. We use global variance (GV) enhancement to improve the perceived quality of the synthesized signals obtained by the proposed and the GMM-based methods. Using just 10 training sentences, our enhanced GB method leads to converted sentences having closer GV values to those of the target and to lower spectral distances at the same time, compared to enhanced version of the GMM-based conversion method. Furthermore, subjective evaluations show that signals produced by the enhanced GB method are perceived as more similar to the target speaker than the enhanced GMM signals, at the expense of a small degradation in the perceived quality.
This paper considers the design of codebooks for low bit-rate speech coders which apply joint seg... more This paper considers the design of codebooks for low bit-rate speech coders which apply joint segmentation-quantization. Two issues are addressed. The first is the design of a codebook from training data when the data vectors vary in size, as dictated by the segmentation process. For this task a modification of the generalized Lloyd algorithm (GLA) is devised. The second issue is assigning indices to the codebook codewords to combat errors due to transmission over a binary symmetric channel (BSC). Two specific useful index assignment (IA) methods are presented
This paper describes a technique for Fractal Image Coding in the Discrete Wavelet Transform (DWT)... more This paper describes a technique for Fractal Image Coding in the Discrete Wavelet Transform (DWT) domain employing variable size range and domain subtrees. The DWT domain is partitioned by means of a top-down quad-tree algorithm. The splitting decision function is implemented by a list ordered with respect to a rate-distortion based partitioning gain. The partitioning and encoding are done at the same pass. Each range subtree has three optional coding modes: zero-tree, fractal prediction by a best fit domain subtree from a pool of the same size and "Intra" scalar quantization of the subtree coefficients. The choice between fractal coding and Intra coding is rate-distortion based.
The Optimized Temporal Decomposition (OTD) technique for Line Spectral Frequencies (LSF) speech e... more The Optimized Temporal Decomposition (OTD) technique for Line Spectral Frequencies (LSF) speech envelope representation, under a MMSE criterion, has been shown to be promising for very low bit rate speech coding for storage and broadcast applications. In order to improve perceptual speech quality, a dynamically weighted OTD (DW-OTD) technique is introduced in this work. It extends the OTD by allowing temporally changing weights, so as to improve the perceived speech quality. Use of Gardner's weighted MSE with DW-OTD is found to reduce the Log Spectral Distance (LSD) measure by 0.3 dB, as compared to OTD. The original OTD algorithm delay and complexity requirements make it inappropriate for real-time speech coding. In this paper we also introduce a modification of this technique, which is suboptimal but suitable for on-line speech coding purposes, with negligible degradation of performance (of only about 0.06 dB in LSD). With the proposed techniques we were able to encode speech spectral envelopes at 300-370 bps at LSD of 2.25-2.1 dB, respectively, with a delay of just 7 frames.
Ah&me-The interpolation formula representation and the kernels associated with the Discrete Fouri... more Ah&me-The interpolation formula representation and the kernels associated with the Discrete Fourier Transform (DFT) approach to the interpolation of periodic signals are obtained by viewing the interpolation process as a filtering operation on a properly defined sequence. This representation provides then the basis for the derivation of upper bounds on the interpolation error involved.
Proceedings of ICASSP '94. IEEE International Conference on Acoustics, Speech and Signal Processing
The paper presents an improved collage theorem, valid for a class of signal mappings called Afine... more The paper presents an improved collage theorem, valid for a class of signal mappings called Afine Blockwise Aueraging (ABA) Mappings. The .4BA structure is exploited to form a bound depending on norms of collage error signals at several resolutions. Compared to previously published collage theorems, the new theorem provides a much tighter bound on the maximum distance between the original signal (real world image) and the decoded attractor, given the distance between the original and the "collage" optimized by the encoder. This is achieved without contractivity constraints on the .4BA mapping parameters. Finally, an encoding method in which one attempts to minimize the upper bound directly is suggested.
2007 IEEE International Conference on Image Processing, 2007
Show-Through is a common occurrence when scanning duplex printed documents. The back-side printin... more Show-Through is a common occurrence when scanning duplex printed documents. The back-side printing shows through the paper, contaminating the front side image. Previous work modeled the problem as a non-linear convolutive mixture of images and offered solutions based on decorrelation. In this work we propose a cleaning process based on a Blind Source Separation approach. We define a cost function incorporating the non-linear mixing model in a mean-squared error term, along with a regularization term based on Total-Variation. We propose a location dependent regularization tradeoff, preserving image edges while removing show-through edges. The images and mixing parameters are estimated using an alternating minimization process, with each stage using only convex optimization methods. The resulting images exhibit significantly lower show-through, both visibly and in objective measures.
We present a generalization of the concept of Convex Sets, based on the Morphological Closing ope... more We present a generalization of the concept of Convex Sets, based on the Morphological Closing operation, and study some of its properties. We also de ne Extreme Points of such Generalized Convex Sets, which generalize the notion of Extreme Points of Convex Sets. Moreover, we apply the above notions to skeleton redundancy removal, and present an algorithm for obtaining an Error-Free Skeleton representation with reduced amount of redundant points, using morphological operations only.
2017 IEEE International Conference on Computer Vision (ICCV), 2017
Local-to global point cloud registration is a challenging task due to the substantial differences... more Local-to global point cloud registration is a challenging task due to the substantial differences between these two types of data, and the different techniques used to acquire them. Global clouds cover large-scale environments and are usually acquired aerially, e.g., 3D modeling of a city using Airborne Laser Scanning (ALS). In contrast, local clouds are often acquired from ground level and at a much smaller range, for example, using Terrestrial Laser Scanning (TLS). The differences are often manifested in point density distribution, occlusions nature, and measurement noise. As a result of these differences, existing point cloud registration approaches, such as keypoint-based registration, tend to fail. We improve upon a different approach, recently proposed, based on converting the global cloud into a viewpoint-based cloud dictionary. We propose a local-toglobal registration method where we replace the dictionary clouds with viewpoint descriptors, consisting of panoramic range-imag...
The Optimized Temporal Decomposition (OTD) technique for Line Spectral Frequencies (LSF) speech e... more The Optimized Temporal Decomposition (OTD) technique for Line Spectral Frequencies (LSF) speech envelope representation, under a MMSE criterion, has been shown to be promising for very low bit rate speech coding for storage and broadcast applications. In order to improve perceptual speech quality, a dynamically weighted OTD (DW-OTD) technique is introduced in this work. It extends the OTD by allowing temporally changing weights, so as to improve the perceived speech quality. Use of Gardner's weighted MSE with DWOTD is found to reduce the Log Spectral Distance (LSD) measure by 0.3 dB, as compared to OTD. The original OTD algorithm delay and complexity requirements make it inappropriate for real-time speech coding. In this paper we also introduce a modification of this technique, which is suboptimal but suitable for on-line speech coding purposes, with negligible degradation of performance (of only about 0.06 dB in LSD). With the proposed techniques we were able to encode speech s...
As 3D point cloud acquisition sensors become increasingly prevalent in urban environments (e.g., ... more As 3D point cloud acquisition sensors become increasingly prevalent in urban environments (e.g., LiDAR sensors for autonomous vehicles), the need arises to find efficient ways to align large amounts of such 3D data, often in real-time. In this work, we propose a novel method for 3D point cloud registration refinement in an urban environment (e.g., between Terrestrial LiDAR Scans - TLS - and Airborne LiDAR Scans - ALS), assuming an initial coarse registration is available. The proposed method is based on estimation of the direction of gravity, wall detection, projection of the point clouds on a perpendicular horizontal plane, and conversion into 2D edge-maps. Then, two methods are considered for alignment between the 2D edge-maps: a 2D variant of the well-known ICP (Iterative Closest Point) algorithm, and Edge-Map Phase-Correlation (EMPC). We demonstrate the usefulness of the proposed methods for registration in this challenging task, where the 2D variant of ICP achieves a meaningful...
We present a new tree-based framework for producing self-dual morphological operators. For any gi... more We present a new tree-based framework for producing self-dual morphological operators. For any given tree representation of images, one can associate a complete inf-semilattice (CISL) in the corresponding tree-representation domain, where the operators can then be derived. We also present a particular case of this general framework, involving a new tree representation, the ExtremaWatershed Tree (EWT). The operators obtained by using the EWT in the above framework behave like classical morphological operators, but in addition are self-dual. Some application examples are provided: pre-processing for OCR and dust & scratch removal algorithms, and image denoising.
Mobile devices are widely used around the world, frequently by people speaking local languages or... more Mobile devices are widely used around the world, frequently by people speaking local languages or dialects that are not well documented. For these languages, it might not be beneficial for commercial companies to develop Automatic Speech Recognition (ASR) systems, so users of these languages cannot utilize voice activation features (often using Keyword Spotting, KWS) of their devices. Standard KWS methods aim to statistically model the generation process of the speech signal, requiring hours of recorded and transcribed speech for training, and therefore are not adequate for limited-data scenarios. In this paper we propose a new KWS method, suitable for limited-data scenarios, which can be easily applied by developers. The proposed method uses a new histogram representation for words, obtained with respect to a pre-trained Gaussian Mixture Model (GMM). Sentences are represented by fixed-length global feature vectors, extracted from the response curves obtained by a word classifier. Word and sentence classifiers are trained using a discriminative approach, which is typically robust to training-set size. The dataset for training the GMM is easy to obtain, since no annotation is required. We compared the proposed system to a Hidden Markov Model (HMM) based system, trained using the same low data-resources conditions as ours, and to a state-of-the-art ASR system, trained using either the limited data scenario, or using many hours of recorded speech. In the limited data situation, our system performs better then both benchmarks in all experiments except for clean speech of children (CSLU dataset), where it performs as good as the HMM. Since the ASR benchmark performs poorly without enough training data, we also trained it without limiting the available data. In this case the ASR benchmark performs better when tested on speech of adults (TED-LIUM dataset of TED lectures) for all noise conditions, and our system performs better when tested on speech of children with low to moderate SNR values. The results demonstrate the advantages of the proposed system, and the conditions under which it performs better.
EURASIP Journal on Audio, Speech, and Music Processing, 2016
The goal of voice conversion is to modify a source speaker's speech to sound as if spoken by a ta... more The goal of voice conversion is to modify a source speaker's speech to sound as if spoken by a target speaker. Common conversion methods are based on Gaussian mixture modeling (GMM). They aim to statistically model the spectral structure of the source and target signals and require relatively large training sets (typically dozens of sentences) to avoid over-fitting. Moreover, they often lead to muffled synthesized output signals, due to excessive smoothing of the spectral envelopes. Mobile applications are characterized with low resources in terms of training data, memory footprint, and computational complexity. As technology advances, computational and memory requirements become less limiting; however, the amount of available training data still presents a great challenge, as a typical mobile user is willing to record himself saying just few sentences. In this paper, we propose the grid-based (GB) conversion method for such low resource environments, which is successfully trained using very few sentences (5-10). The GB approach is based on sequential Bayesian tracking, by which the conversion process is expressed as a sequential estimation problem of tracking the target spectrum based on the observed source spectrum. The converted Mel frequency cepstrum coefficient (MFCC) vectors are sequentially evaluated using a weighted sum of the target training vectors used as grid points. The training process includes simple computations of Euclidian distances between the training vectors and is easily performed even in cases of very small training sets. We use global variance (GV) enhancement to improve the perceived quality of the synthesized signals obtained by the proposed and the GMM-based methods. Using just 10 training sentences, our enhanced GB method leads to converted sentences having closer GV values to those of the target and to lower spectral distances at the same time, compared to enhanced version of the GMM-based conversion method. Furthermore, subjective evaluations show that signals produced by the enhanced GB method are perceived as more similar to the target speaker than the enhanced GMM signals, at the expense of a small degradation in the perceived quality.
This paper considers the design of codebooks for low bit-rate speech coders which apply joint seg... more This paper considers the design of codebooks for low bit-rate speech coders which apply joint segmentation-quantization. Two issues are addressed. The first is the design of a codebook from training data when the data vectors vary in size, as dictated by the segmentation process. For this task a modification of the generalized Lloyd algorithm (GLA) is devised. The second issue is assigning indices to the codebook codewords to combat errors due to transmission over a binary symmetric channel (BSC). Two specific useful index assignment (IA) methods are presented
This paper describes a technique for Fractal Image Coding in the Discrete Wavelet Transform (DWT)... more This paper describes a technique for Fractal Image Coding in the Discrete Wavelet Transform (DWT) domain employing variable size range and domain subtrees. The DWT domain is partitioned by means of a top-down quad-tree algorithm. The splitting decision function is implemented by a list ordered with respect to a rate-distortion based partitioning gain. The partitioning and encoding are done at the same pass. Each range subtree has three optional coding modes: zero-tree, fractal prediction by a best fit domain subtree from a pool of the same size and "Intra" scalar quantization of the subtree coefficients. The choice between fractal coding and Intra coding is rate-distortion based.
The Optimized Temporal Decomposition (OTD) technique for Line Spectral Frequencies (LSF) speech e... more The Optimized Temporal Decomposition (OTD) technique for Line Spectral Frequencies (LSF) speech envelope representation, under a MMSE criterion, has been shown to be promising for very low bit rate speech coding for storage and broadcast applications. In order to improve perceptual speech quality, a dynamically weighted OTD (DW-OTD) technique is introduced in this work. It extends the OTD by allowing temporally changing weights, so as to improve the perceived speech quality. Use of Gardner's weighted MSE with DW-OTD is found to reduce the Log Spectral Distance (LSD) measure by 0.3 dB, as compared to OTD. The original OTD algorithm delay and complexity requirements make it inappropriate for real-time speech coding. In this paper we also introduce a modification of this technique, which is suboptimal but suitable for on-line speech coding purposes, with negligible degradation of performance (of only about 0.06 dB in LSD). With the proposed techniques we were able to encode speech spectral envelopes at 300-370 bps at LSD of 2.25-2.1 dB, respectively, with a delay of just 7 frames.
Ah&me-The interpolation formula representation and the kernels associated with the Discrete Fouri... more Ah&me-The interpolation formula representation and the kernels associated with the Discrete Fourier Transform (DFT) approach to the interpolation of periodic signals are obtained by viewing the interpolation process as a filtering operation on a properly defined sequence. This representation provides then the basis for the derivation of upper bounds on the interpolation error involved.
Uploads
Papers by David Malah