The rise of generative AI has democratized media creation, bringing huge promise but also possible perils. While this may seem like a new problem, the generation and manipulation of media has a long history that predates the current AI boom. I'll discuss key insights from our multi-year analysis of content that people shared online. Looking at manipulations such as deepfakes and cheapfakes, as well as misleading contextual manipulations, I'll reveal surprising statistics that challenge common assumptions about the most prevalent types of problematic media. I'll then explore mitigation strategies, including ways to improve information literacy tools, the opportunities and limitations of using AI to detect manipulated content, and how provenance methods paired with AI can help address out-of-context manipulations. Finally, I'll introduce an AI-based tool that can provide additional context for the media we encounter online every day.

Despite the recent attention to DeepFakes, one of the most prevalent ways to mislead audiences on social media is the use of unaltered images in a new but false context. We propose a new method that automatically highlights out-of-context image and text pairs, for assisting fact-checkers. Our key insight is to leverage the grounding of images with text to distinguish out-of-context scenarios that cannot be disambiguated with language alone. We propose a self-supervised training strategy where we only need a set of captioned images. At train time, our method learns to selectively align individual objects in an image with textual claims, without explicit supervision. At test time, we check if both captions correspond to the same object(s) in the image but are semantically different, which allows us to make fairly accurate out-of-context predictions. Our method achieves 85% out-of-context detection accuracy. To facilitate benchmarking of this task, we create a large-scale dataset of 200K images with 450K textual captions from a variety of news websites, blogs, and social media posts

In this paper, we present a video-based learning framework for animating personalized 3D talking faces from audio. We introduce two training-time data normalizations that significantly improve data sample efficiency. First, we isolate and represent faces in a normalized space that decouples 3D geometry, head pose, and texture. This decomposes the prediction problem into regressions over the 3D face shape and the corresponding 2D texture atlas. Second, we leverage facial symmetry and approximate albedo constancy of skin to isolate and remove spatio-temporal lighting variations. Together, these normalizations allow simple networks to generate high fidelity lip-sync videos under novel ambient illumination while training with just a single speaker-specific video. Further, to stabilize temporal dynamics, we introduce an auto-regressive approach that conditions the model on its previous visual state. Human ratings and objective metrics demonstrate that our method outperforms contemporary state-of-the-art audio-driven video reenactment benchmarks in terms of realism, lip-sync and visual quality scores. We illustrate several applications enabled by our framework.

In this work we propose a model that can manipulate individual visual attributes of objects in a real scene using examples of how respective attribute manipulations affect the output of a simulation. As an example, we train our model to manipulate the expression of a human face using nonphotorealistic 3D renders of a face with varied expression. Our model manages to preserve all other visual attributes of a real face, such as head orientation, even though this and other attributes are not labeled in either real or synthetic domain. Since our model learns to manipulate a specific property in isolation using only "synthetic demonstrations" of such manipulations without explicitly provided labels, it can be applied to shape, texture, lighting, and other properties that are difficult to measure or represent as real-valued vectors. We measure the degree to which our model preserves other attributes of a real image when a single specific attribute is manipulated. We use digit datasets to analyze how discrepancy in attribute distributions affects the performance of our model, and demonstrate results in a far more difficult setting: learning to manipulate real human faces using nonphotorealistic 3D renders.

We propose a method for multi-person detection and 2-D pose estimation that achieves state-of-art results on the challenging COCO keypoints task. It is a simple, yet powerful, top-down approach consisting of two stages. In the first stage, we predict the location and scale of boxes which are likely to contain people, for this we use the Faster RCNN detector. In the second stage, we estimate the keypoints of the person potentially contained in each proposed bounding box. For each keypoint type we predict dense heatmaps and offsets using a fully convolutional ResNet. To combine these outputs we introduce a novel aggregation procedure to obtain highly localized keypoint predictions. We also use a novel form of keypoint-based Non-Maximum-Suppression (NMS), instead of the cruder box-level NMS, and a novel form of keypoint-based confidence score estimation, instead of box-level scoring. Trained on COCO data alone, our final system achieves average precision of 0.649 on the COCO test-dev set and the 0.643 test-standard sets, outperforming the winner of the 2016 COCO keypoints challenge and other recent state-of-art. Further, by using additional in-house labeled data we obtain an even higher average precision of 0.685 on the test-dev set and 0.673 on the test-standard set, more than 5% absolute improvement compared to the previous best performing method on the same dataset.

Marker-less motion capture has seen great progress, but most state-of-the-art approaches fail to reliably track articulated human body motion with a very low number of cameras, let alone when applied in outdoor scenes with general background. In this paper, we propose a method for accurate marker-less capture of articulated skeleton motion of several subjects in general scenes, indoors and outdoors, even from input filmed with as few as two cameras. The new algorithm combines the strengths of a discriminative image-based joint detection method with a model-based generative motion tracking algorithm through an unified pose optimization energy. The discriminative part-based pose detection method is implemented using Convolutional Networks (ConvNet) and estimates unary potentials for each joint of a kinematic skeleton model. These unary potentials serve as the basis of a probabilistic extraction of pose constraints for tracking by using weighted sampling from a pose posterior that is guided by the model. In the final energy, we combine these constraints with an appearance-based model-to-image similarity term. Poses can be computed very efficiently using iterative local optimization, since joint detection with a trained ConvNet is fast, and since our formulation yields a combined pose estimation energy with analytic derivatives. In combination, this enables to track full articulated joint angles at state-of-the-art accuracy and temporal stability with a very low number of cameras. Our method is efficient and lends itself to implementation on parallel computing hardware, such as GPUs. We test our method extensively and show its advantages over related work on many indoor and outdoor data sets captured by ourselves, as well as data sets made available to the community by other research labs. The availability of good evaluation data sets is paramount for scientific progress, and many existing test data sets focus on controlled indoor settings, do not feature much variety in the scenes, and often lack a large corpus of data with ground truth annotation. We therefore further contribute with a new extensive test data set called MPI-MARCOnI for indoor and outdoor marker-less motion capture that features 12 scenes of varying complexity and varying camera count, and that features ground truth reference data from different modalities, ranging from manual joint annotations to marker-based motion capture results. Our new method is tested on these data, and the data set will be made available to the community.

We present a novel method for accurate marker-less capture of articulated skeleton motion of several subjects in general scenes, indoors and outdoors, even from input filmed with as few as two cameras. Our approach unites a discriminative image-based joint detection method with a model-based generative motion tracking algorithm through a combined pose optimization energy. The discriminative part-based pose detection method, implemented using Convolutional Networks (ConvNet), estimates unary potentials for each joint of a kinematic skeleton model. These unary potentials are used to probabilistically extract pose constraints for tracking by using weighted sampling from a pose posterior guided by the model. In the final energy, these constraints are combined with an appearance-based model-to-image similarity term. Poses can be computed very efficiently using iterative local optimization, as ConvNet detection is fast, and our formulation yields a combined pose estimation energy with analytic derivatives. In combination, this enables to track full articulated joint angles at state-of-the-art accuracy and temporal stability with a very low number of cameras.

Recent state-of-the-art performance on human-body pose estimation has been achieved with Deep Convolutional Networks (ConvNets). Traditional ConvNet architectures include pooling and sub-sampling layers which reduce computational requirements, introduce invariance and prevent over-training. These benefits of pooling come at the cost of reduced localization accuracy. We introduce a novel architecture which includes an efficient ‘position refinement’ model that is trained to estimate the joint offset location within a small region of the image. This refinement model is jointly trained in cascade with a state-of-the-art ConvNet model [21] to achieve improved accuracy in human joint location estimation. We show that the variance of our detector approaches the variance of human annotations on the FLIC [20] dataset and outperforms all existing approaches on the MPII-human-pose dataset [1].

Dance is a dynamic art form that reflects a wide range of cultural diversity and individuality. With the advancement of motion-capture technology combined with crowd-sourcing and machine learning algorithms, we explore the complex relationship between perceived dance quality/dancer's gender and dance movements/music respectively. As a feasibility study, we construct a computational framework for an analysis-synthesis-feedback loop using a novel multimedia dance-music texture representation. Furthermore, we integrate crowd-sourcing, music and motion-capture data, and machine learning-based methods for dance segmentation, analysis and synthesis of new dancers. A quantitative validation of this framework on a motion-capture dataset of 172 dancers evaluated by more than 400 independent on-line raters demonstrates significant correlation between human perception and the algorithmically intended dance quality or gender of synthesized dancers. The technology illustrated in this work has a high potential to advance the multimedia entertainment industry via dancing with Turks.

We introduce a real-time and calibration-free facial performance capture framework based on a sensor with video and depth input. In this framework, we develop an adaptive PCA model using shape correctives that adjust on-the-fly to the actor's expressions through incremental PCA-based learning. Since the fitting of the adaptive model progressively improves during the performance, we do not require an extra capture or training session to build this model. As a result, the system is highly deployable and easy to use: it can faithfully track any individual, starting from just a single face scan of the subject in a neutral pose. Like many real-time methods, we use a linear subspace to cope with incomplete input data and fast motion. To boost the training of our tracking model with reliable samples, we use a well-trained 2D facial feature tracker on the input video and an efficient mesh deformation algorithm to snap the result of the previous step to high frequency details in visible depth map regions. We show that the combination of dense depth maps and texture features around eyes and lips is essential in capturing natural dialogues and nuanced actor-specific emotions. We demonstrate that using an adaptive PCA model not only improves the fitting accuracy for tracking but also increases the expressiveness of the retargeted character.

Supervised methods for learning an embedding aim to map high-dimensional images to a space in which perceptually similar observations have high measurable similarity. Most approaches rely on binary similarity, typically defined by class membership where labels are expensive to obtain and/or difficult to define. In this paper we propose crowd-sourcing similar images by soliciting human imitations. We exploit temporal coherence in video to generate additional pairwise graded similarities between the user-contributed imitations. We introduce two methods for learning nonlinear, invariant mappings that exploit graded similarities. We learn a model that is highly effective at matching people in similar pose. It exhibits remarkable invariance to identity, clothing, background, lighting, shift and scale.

This paper shows how “Body Motion Signature Analysis” - a new “soft-biometrics” technique - can be used for identity verification. It is able to extract motion features from the upper body of people and estimates so called “super-features” for input to a classifier. We demonstrate how this new technique can be used to identify people just based on their motion, or it can be used to significantly improve “hard-biometrics” techniques. For example, face verification achieves on this domain 6.45% Equal Error Rate (EER), and the combined verification performance of motion features and face reduces the error to 4.96% using an adaptive score-level integration method. The more ambiguous motion-only performance is 17.1% EER.

This paper tackles the complex problem of visually matching people in similar pose but with different clothes, background, and other appearance changes. We achieve this with a novel method for learning a nonlinear embedding based on several extensions to the Neighborhood Component Analysis (NCA) framework. Our method is convolutional, enabling it to scale to realistically-sized images. By cheaply labeling the head and hands in large video databases through Amazon Mechanical Turk (a crowd-sourcing service), we can use the task of localizing the head and hands as a proxy for determining body pose. We apply our method to challenging real-world data and show that it can generalize beyond hand localization to infer a more general notion of body pose. We evaluate our method quantitatively against other embedding methods. We also demonstrate that real-world performance can be improved through the use of synthetic data.

The literature currently provides two ways to establish point correspondences between images with moving objects. On one side, there are energy minimization methods that yield very accurate, dense flow fields, but fail as displacements get too large. On the other side, there is descriptor matching that allows for large displacements, but correspondences are very sparse, have limited accuracy, and due to missing regularity constraints there are many outliers. In this paper we propose a method that can combine the advantages of both matching strategies. A region hierarchy is established for both images. Descriptor matching on these regions provides a sparse set of hypotheses for correspondences. These are integrated into a variational approach and guide the local optimization to large displacement solutions. The variational optimization selects among the hypotheses and provides dense and subpixel accurate estimates, making use of geometric constraints and all available image information.

We show how an SVM based acoustic speaker verification system can be significantly improved in incorporating new visual features that capture the speaker's ldquobody language.rdquo We apply this system to many hours of Internet videos and TV broadcasts of politicians and other public figures. Our data ranges from current and former US election candidates to the Queen of England, the President of France, and the Pope, while giving speeches.

This paper describes methods for recovering time-varying shape and motion of non-rigid 3D objects from uncalibrated 2D point tracks. For example, given a video recording of a talking person, we would like to estimate the 3D shape of the face at each instant, and learn a model of facial deformation. Time-varying shape is modeled as a rigid transformation combined with a non-rigid deformation. Reconstruction is ill-posed if arbitrary deformations are allowed, and thus additional assumptions about deformations are required. We first suggest restricting shapes to lie within a low-dimensional subspace, and describe estimation algorithms. However, this restriction alone is insufficient to constrain reconstruction. To address these problems, we propose a reconstruction method using a Probabilistic Principal Components Analysis (PPCA) shape model, and an estimation algorithm that simultaneously estimates 3D shape and motion for each instant, learns the PPCA model parameters, and robustly fills-in missing data points. We then extend the model to model temporal dynamics in object shape, allowing the algorithm to robustly handle severe cases of missing data.

This paper presents an algorithm for synthesis of human motion in specified styles. We use a theory of movement observation (Laban Movement Analysis) to describe movement styles as points in a multi-dimensional perceptual space. We cast the task of learning to synthesize desired movement styles as a regression problem: sequences generated via space-time interpolation of motion capture data are used to learn a nonlinear mapping between animation parameters and movement styles in perceptual space. We demonstrate that the learned model can apply a variety of motion styles to pre-recorded motion sequences and it can extrapolate styles not originally included in the training data.

Motion capture-based facial animation has recently gained popularity in many applications, such as movies, video games, and human-computer interface designs. With the use of sophisticated facial motions from a human performer, animated characters are far more lively and convincing. However, editing motion data is difficult, limiting the potential of reusing the motion data for different tasks. To address this problem, statistical techniques have been applied to learn models of the facial motion in order to derive new motions based on the existing data. Most existing research focuses on audio-to-visual mapping and reordering of words, or on photo-realistically matching the synthesized face to the original performer. Little attention has been paid to modifying and controlling facial expression, or to mapping expressive motion onto other 3D characters.This article describes a method for creating expressive facial animation by extracting information from the expression axis of a speech performance. First, a statistical model for factoring the expression and visual speech is learned from video. This model can be used to analyze the facial expression of a new performance or modify the facial expressions of an existing performance. With the addition of this analysis of the facial expression, the facial motion can be more effectively retargeted to another 3D face model. The blendshape retargeting technique is extended to include subsets of morph targets that belong to different facial expression groups. The proportion of each subset included in a final animation is weighted according to the expression information. The resulting animation conveys much more emotion than if only the motion vectors were used for retargeting. Finally, since head motion is very important in adding liveness to facial animation, we introduces an audio-driven synthesis technique for generating new head motion.

We describe a method for using a database of recorded speech and captured motion to create an animated conversational character. People's utterances are composed of short, clearly-delimited phrases; in each phrase, gesture and speech go together meaningfully and synchronize at a common point of maximum emphasis. We develop tools for collecting and managing performance data that exploit this structure. The tools help create scripts for performers, help annotate and segment performance data, and structure specific messages for characters to use within application contexts. Our animations then reproduce this structure. They recombine motion samples with new speech samples to recreate coherent phrases, and blend segments of speech and motion together phrase-by-phrase into extended utterances. By framing problems for utterance generation and synthesis so that they can draw closely on a talented performance, our techniques support the rapid construction of animated characters with rich and appropriate expression.

This paper presents an algorithm for learning the time-varying shape of a non-rigid 3D object from uncalibrated 2D tracking data. We model shape motion as a rigid component (rotation and translation) combined with a non-rigid deformation. Reconstruction is ill-posed if arbitrary deforma- tions are allowed. We constrain the problem by assuming that the object shape at each time instant is drawn from a Gaussian distribution. Based on this assumption, the algorithm simultaneously estimates 3D shape and motion for each time frame, learns the parameters of the Gaussian, and robustly fills-in missing data points. We then extend the algorithm to model temporal smoothness in object shape, thus allowing it to handle severe cases of missing data.

Animation of facial speech and expressions has experienced increased attention recently. Most current research focuses on techniques for capturing, synthesizing, and retargeting facial expressions. Little attention has been paid to the problem of controlling and modifying the expression itself. We present techniques that separate video data into expressive features and underlying content. This allows, for example, a sequence originally recorded with a happy expression to be modified so that the speaker appears to be speaking with an angry or neutral expression. Although the expression has been modified, the new sequences maintain the same visual speech content as the original sequence. The facial expression space that allows these transformations is learned with the aid of a factorization model.

In this paper, we present a technique we call "cartoon capture and retargeting" which we use to track the motion from traditionally animated cartoons and retarget it onto 3-D models, 2-D drawings, and photographs. By using animation as the source, we can produce new animations that are expressive, exaggerated or non-realistic.Cartoon capture transforms a digitized cartoon into a cartoon motion representation. Using a combination of affine transformation and key-shape interpolation, cartoon capture tracks non-rigid shape changes in cartoon layers. Cartoon retargeting translates this information into different output media. The result is an animation with a new look but with the movement of the original cartoon.

We discuss a method for creating animations that allows the animator to sketch an animation by setting a small number of keyframes on a fraction of the possible degrees of freedom. Motion capture data is then used to enhance the animation. Detail is added to degrees of freedom that were keyframed, a process we call texturing. Degrees of freedom that were not keyframed are synthesized. The method takes advantage of the fact that joint motions of an articulated figure are often correlated, so that given an incomplete data set, the missing degrees of freedom can be predicted from those that are present.

This paper presents a novel solution for flow-based tracking and 3D reconstruction of deforming objects in monocular image sequences. A non-rigid 3D object undergoing rotation and deformation can be effectively approximated using a linear combination of 3D basis shapes. This puts a bound on the rank of the tracking matrix. The rank constraint is used to achieve robust and precise low-level optical flow estimation without prior knowledge of the 3D shape of the object. The bound on the rank is also exploited to handle occlusion at the tracking level leading to the possibility of recovering the complete trajectories of occluded/disoccluded points. Following the same low-rank principle, the resulting flow matrix can be factored to get the 3D pose, configuration coefficients, and 3D basis shapes. The flow matrix is factored in an iterative manner, looping between solving for pose, configuration, and basis shapes. The flow-based tracking is applied to several video sequences and provides the input to the 3D non-rigid reconstruction task. Additional results on synthetic data and comparisons to ground truth complete the experiments.

We describe a method for synthesizing joint angle and translation data based on the information in motion capture data. The synthetic data is realistic not only in that it resembles the original training data, but in that it has random variations that are statistically similar to what one would find in repeated measurements of the motion. To achieve this result, the training data is broken into frequency bands using a wavelet decomposition, and the information in these bands is used to create the synthetic data one frequency band at a time. The method takes into account the fact that there are correlations among numerous features of the data. For example, a point characterized by a particular time and frequency band will depend upon points close to it in time in other frequency bands. Such correlations are modeled with a kernel-based representation of the joint probability distributions of the features. The data is synthesized by sampling from these densities and improving the results using a new iterative maximization technique. We have applied this technique to the synthesis of joint angle and translation data of a wallaby hopping on a treadmill. The synthetic data was used to animate characters that have limbs proportional to the wallaby.

The paper addresses the problem of recovering 3D non-rigid shape models from image sequences. For example, given a video recording of a talking person, we would like to estimate a 3D model of the lips and the full face and its internal modes of variation. Many solutions that recover 3D shape from 2D image sequences have been proposed; these so-called structure-from-motion techniques usually assume that the 3D object is rigid. For example, C. Tomasi and T. Kanades' (1992) factorization technique is based on a rigid shape matrix, which produces a tracking matrix of rank 3 under orthographic projection. We propose a novel technique based on a non-rigid model, where the 3D shape in each frame is a linear combination of a set of basis shapes. Under this model, the tracking matrix is of higher rank, and can be factored in a three-step process to yield pose, configuration and shape. To the best of our knowledge, this is the first model free approach that can recover from single-view video sequences nonrigid shape models. We demonstrate this new algorithm on several video sequences. We were able to recover 3D non-rigid human face and animal models with high accuracy.

This paper demonstrates a new visual motion estimation technique that is able to recover high degree-of-freedom articulated human body configurations in complex video sequences. We introduce the use of a novel mathematical technique, the product of exponential maps and twist motions, and its integration into a differential motion estimation. This results in solving simple linear systems, and enables us to recover robustly the kinematic degrees-of-freedom in noise and complex self occluded configurations. We demonstrate this on several image sequences of people doing articulated full body movements, and visualize the results in re-animating an artificial 3D human model. We are also able to recover and re-animate the famous movements of Eadweard Muybridge's motion studies from the last century. To the best of our knowledge, this is the first computer vision based system that is able to process such challenging footage and recover complex motions with such high accuracy.

This paper describes a probabilistic decomposition of human dynamics at multiple abstractions, and shows how to propagate hypotheses across space, time, and abstraction levels. Recognition in this framework is the succession of very general low level grouping mechanisms to increased specific and learned model based grouping techniques at higher levels. Hard decision thresholds are delayed and resolved by higher level statistical models and temporal context. Low-level primitives are areas of coherent motion found by EM clustering, mid-level categories are simple movements represented by dynamical systems, and high-level complex gestures are represented by Hidden Markov Models as successive phases of ample movements. We show how such a representation can be learned from training data, and apply It to the example of human gait recognition.

Article Video Rewrite: driving visual speech with audio Share on Authors: Christoph Bregler Interval Research Corporation, 1801 Page Mill Road, Building C, Palo Alto, CA Interval Research Corporation, 1801 Page Mill Road, Building C, Palo Alto, CAView Profile , Michele Covell Interval Research Corporation, 1801 Page Mill Road, Building C, Palo Alto, CA Interval Research Corporation, 1801 Page Mill Road, Building C, Palo Alto, CAView Profile , Malcolm Slaney Interval Research Corporation, 1801 Page Mill Road, Building C, Palo Alto, CA Interval Research Corporation, 1801 Page Mill Road, Building C, Palo Alto, CAView Profile Authors Info & Claims SIGGRAPH '97: Proceedings of the 24th annual conference on Computer graphics and interactive techniquesAugust 1997 Pages 353–360https://doi.org/10.1145/258734.258880Published:03 August 1997 385citation1,950DownloadsMetricsTotal Citations385Total Downloads1,950Last 12 Months132Last 6 weeks23 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteGet Access

Eigen-points places control points onto unmarked images. The control points are the image locations corresponding to fiduciary points on an object. For example, we might designate ten points on the outside boundary of the lip as fiduciary points on a face. Then, the control points mark the image locations where those points on the outside lip boundary appear. The control-point locations are estimated using a coupled manifold model, which describes the joint variation of the image appearance and the control-point location. The paper first discusses why this problem is interesting and then reviews previous approaches to placing control points on images of deformable objects. It also outlines our approach to placing control points automatically. Finally, some results from our analysis are presented.

Retrieving images from very large collections, using image content as a key, is becoming an important problem. Finding objects in image databases is a big challenge in the field. The paper describes our approach to object recognition, which is distinguished by: a rich involvement of early visual primitives, including color and texture; hierarchical grouping and learning strategies in the classification process; the ability to deal with rather general objects in uncontrolled configurations and contexts. We illustrate these properties with three case studies: one demonstrating the use of color and texture descriptors; one learning scenery concepts using grouped features; and one demonstrating a possible application domain in detecting naked people in a scene.

A technique for representing and learning smooth nonlinear manifolds is presented and applied to several lip reading tasks. Given a set of points drawn from a smooth manifold in an abstract feature space, the technique is capable of determining the structure of the surface and of finding the closest manifold point to a given query point. We use this technique to learn the "space of lips" in a visual speech recognition task. The learned manifold is used for tracking and extracting the lips, for interpolating between frames in an image sequence and for providing features for recognition. We describe a system based on hidden Markov models and this learned lip manifold that significantly improves the performance of acoustic speech recognizers in degraded environments. We also present preliminary results on a purely visual lip reader.>

We improve the performance of a hybrid connectionist speech recognition system by incorporating visual information about the corresponding lip movements. Specifically, we investigate the benefits of adding visual features in the presence of additive noise and crosstalk (cocktail party effect). Our study extends our previous experiments by using a new visual front end, and an alternative architecture for combining the visual and acoustic information. Furthermore, we have extended our recognizer to a multi-speaker, connected letters recognizer. Our results show a significant improvement for the combined architecture (acoustic and visual information) over just the acoustic system in the presence of additive noise and crosstalk.>

The problem of interpolating between specified images in an image sequence is a simple, but important task in model-based vision. We describe an approach based on the abstract task of "manifold learning" and present results on both synthetic and real image se(cid:173) quences. This problem arose in the development of a combined lip-reading and speech recognition system.