Multi Pattern Dynamic Time Warping For Automatic Speech Recognition

We are addressing the problem of jointly using multiple noisy speech patterns for automatic speech recognition (ASR), given that they come from the same class. If the user utters a word K times, the ASR system should try to use the information content in all the K patterns of the word simultaneously and improve its speech recognition accuracy compared to that of the single pattern based speech recognition. T address this problem, recently we proposed a Multi Pattern Dynamic Time Warping (MPDTW) algorithm to align the K patterns by finding the least distortion path between them. A Constrained Multi Pattern Viterbi algorithm was used on this aligned path for isolated word recognition (IWR). In this paper, we explore the possibility of using only the MPDTW algorithm for IWR. We also study the properties of the MPDTW algorithm. We show that using only 2 noisy test patterns (10 percent burst noise at -5 dB SNR) reduces the noisy speech recognition error rate by 37.66 percent when compared to the single pattern recognition using the Dynamic Time Warping algorithm.

Joint evaluation of multiple speech patterns for speech recognition and training

We are addressing the novel problem of jointly evaluating multiple speech patterns for automatic speech recognition and training. We propose solutions based on both the non-parametric dynamic time warping (DTW) algorithm, and the parametric hidden Markov model (HMM). We show that a hybrid approach is quite effective for the application of noisy speech recognition. We extend the concept to HMM training wherein some patterns may be noisy or distorted. Utilizing the concept of ``virtual pattern'' developed for joint evaluation, we propose selective iterative training of HMMs. Evaluating these algorithms for burst/transient noisy speech and isolated word recognition, significant improvement in recognition accuracy is obtained using the new algorithms over those which do not utilize the joint evaluation strategy.

Online Unsupervised Pattern Discovery in Speech Using Parallelization

Segmental dynamic time warping (DTW) has been demonstrated to be a useful technique for finding acoustic similarity scores between segments of two speech utterances. Due to its high computational requirements, it had to be computed in an offline manner, limiting the applications of the technique. In this paper, we present results of parallelization of this task by distributing the workload in either a static or dynamic way on an 8-processor cluster and discuss the trade-offs among different distribution schemes. We show that online unsupervised pattern discovery using segmental DTW is plausible with as low as 8 processors. This brings the task within reach of today's general purpose multi-core servers. We also show results on a 32-processor system, and discuss factors affecting scalability of our methods.

Fast Likelihood Computation in Speech Recognition using Matrices

Acoustic modeling using mixtures of multivariate Gaussians is the prevalent approach for many speech processing problems. Computing likelihoods against a large set of Gaussians is required as a part of many speech processing systems and it is the computationally dominant phase for Large Vocabulary Continuous Speech Recognition (LVCSR) systems. We express the likelihood computation as a multiplication of matrices representing augmented feature vectors and Gaussian parameters. The computational gain of this approach over traditional methods is by exploiting the structure of these matrices and efficient implementation of their multiplication. In particular, we explore direct low-rank approximation of the Gaussian parameter matrix and indirect derivation of low-rank factors of the Gaussian parameter matrix by optimum approximation of the likelihood matrix. We show that both the methods lead to similar speedups but the latter leads to far lesser impact on the recognition accuracy. Experiments on 1,138 work vocabulary RM1 task and 6,224 word vocabulary TIMIT task using Sphinx 3.7 system show that, for a typical case the matrix multiplication based approach leads to overall speedup of 46 % on RM1 task and 115 % for TIMIT task. Our low-rank approximation methods provide a way for trading off recognition accuracy for a further increase in computational performance extending overall speedups up to 61 % for RM1 and 119 % for TIMIT for an increase of word error rate (WER) from 3.2 to 3.5 % for RM1 and for no increase in WER for TIMIT. We also express pairwise Euclidean distance computation phase in Dynamic Time Warping (DTW) in terms of matrix multiplication leading to saving of approximately of computational operations. In our experiments using efficient implementation of matrix multiplication, this leads to a speedup of 5.6 in computing the pairwise Euclidean distances and overall speedup up to 3.25 for DTW.

Dynamic Space Warping Of Strokes For Recognition Of Online Handwritten Characters

This paper suggests a scheme for classifying online handwritten characters, based on dynamic space warping of strokes within the characters. A method for segmenting components into strokes using velocity profiles is proposed. Each stroke is a simple arbitrary shape and is encoded using three attributes. Correspondence between various strokes is established using Dynamic Space Warping. A distance measure which reliably differentiates between two corresponding simple shapes (strokes) has been formulated thus obtaining a perceptual distance measure between any two characters. Tests indicate an accuracy of over 85% on two different datasets of characters.

Unrestricted Kannada online handwritten akshara recognition using SDTW

In this paper, we present an unrestricted Kannada online handwritten character recognizer which is viable for real time applications. It handles Kannada and Indo-Arabic numerals, punctuation marks and special symbols like $, &, # etc, apart from all the aksharas of the Kannada script. The dataset used has handwriting of 69 people from four different locations, making the recognition writer independent. It was found that for the DTW classifier, using smoothed first derivatives as features, enhanced the performance to 89% as compared to preprocessed co-ordinates which gave 85%, but was too inefficient in terms of time. To overcome this, we used Statistical Dynamic Time Warping (SDTW) and achieved 46 times faster classification with comparable accuracy i.e. 88%, making it fast enough for practical applications. The accuracies reported are raw symbol recognition results from the classifier. Thus, there is good scope of improvement in actual applications. Where domain constraints such as fixed vocabulary, language models and post processing can be employed. A working demo is also available on tablet PC for recognition of Kannada words.

Joint decoding of multiple speech patterns for robust speech recognition

We are addressing a new problem of improving automatic speech recognition performance, given multiple utterances of patterns from the same class. We have formulated the problem of jointly decoding K multiple patterns given a single Hidden Markov Model. It is shown that such a solution is possible by aligning the K patterns using the proposed Multi Pattern Dynamic Time Warping algorithm followed by the Constrained Multi Pattern Viterbi Algorithm The new formulation is tested in the context of speaker independent isolated word recognition for both clean and noisy patterns. When 10 percent of speech is affected by a burst noise at -5 dB Signal to Noise Ratio (local), it is shown that joint decoding using only two noisy patterns reduces the noisy speech recognition error rate to about 51 percent, when compared to the single pattern decoding using the Viterbi Algorithm. In contrast a simple maximization of individual pattern likelihoods, provides only about 7 percent reduction in error rate.

Learning Feature Trajectories Using Gabor Filter Bank for Human Activity Segmentation and Recognition

We describe a novel method for human activity segmentation and interpretation in surveillance applications based on Gabor filter-bank features. A complex human activity is modeled as a sequence of elementary human actions like walking, running, jogging, boxing, hand-waving etc. Since human silhouette can be modeled by a set of rectangles, the elementary human actions can be modeled as a sequence of a set of rectangles with different orientations and scales. The activity segmentation is based on Gabor filter-bank features and normalized spectral clustering. The feature trajectories of an action category are learnt from training example videos using dynamic time warping. The combined segmentation and the recognition processes are very efficient as both the algorithms share the same framework and Gabor features computed for the former can be used for the later. We have also proposed a simple shadow detection technique to extract good silhouette which is necessary for good accuracy of an action recognition technique.

Comparison of HMM and SDTW for Tamil handwritten character recognition

In this paper, we compare the experimental results for Tamil online handwritten character recognition using HMM and Statistical Dynamic Time Warping (SDTW) as classifiers. HMM was used for a 156-class problem. Different feature sets and values for the HMM states & mixtures were tried and the best combination was found to be 16 states & 14 mixtures, giving an accuracy of 85%. The features used in this combination were retained and a SDTW model with 20 states and single Gaussian was used as classifier. Also, the symbol set was increased to include numerals, punctuation marks and special symbols like $, & and #, taking the number of classes to 188. It was found that, with a small addition to the feature set, this simple SDTW classifier performed on par with the more complicated HMM model, giving an accuracy of 84%. Mixture density estimation computations was reduced by 11 times. The recognition is writer independent, as the dataset used is quite large, with a variety of handwriting styles.

Performance enhancement of online handwritten Tamil symbol recognition with reevaluation techniques

In this article, we aim at reducing the error rate of the online Tamil symbol recognition system by employing multiple experts to reevaluate certain decisions of the primary support vector machine classifier. Motivated by the relatively high percentage of occurrence of base consonants in the script, a reevaluation technique has been proposed to correct any ambiguities arising in the base consonants. Secondly, a dynamic time-warping method is proposed to automatically extract the discriminative regions for each set of confused characters. Class-specific features derived from these regions aid in reducing the degree of confusion. Thirdly, statistics of specific features are proposed for resolving any confusions in vowel modifiers. The reevaluation approaches are tested on two databases (a) the isolated Tamil symbols in the IWFHR test set, and (b) the symbols segmented from a set of 10,000 Tamil words. The recognition rate of the isolated test symbols of the IWFHR database improves by 1.9 %. For the word database, the incorporation of the reevaluation step improves the symbol recognition rate by 3.5 % (from 88.4 to 91.9 %). This, in turn, boosts the word recognition rate by 11.9 % (from 65.0 to 76.9 %). The reduction in the word error rate has been achieved using a generic approach, without the incorporation of language models.

Online Handwritten Kannada Word Recognizer with Unrestricted Vocabulary

In this paper, we propose a novel heuristic approach to segment recognizable symbols from online Kannada word data and perform recognition of the entire word. Two different estimates of first derivative are extracted from the preprocessed stroke groups and used as features for classification. Estimate 2 proved better resulting in 88% accuracy, which is 3% more than that achieved with estimate 1. Classification is performed by statistical dynamic space warping (SDSW) classifier which uses X, Y co-ordinates and their first derivatives as features. Classifier is trained with data from 40 writers. 295 classes are handled covering Kannada aksharas, with Kannada numerals, Indo-Arabic numerals, punctuations and other special symbols like $ and #. Classification accuracies obtained are 88% at the akshara level and 80% at the word level, which shows the scope for further improvement in segmentation algorithm

Traffic Profiling for Efficient Network Resource Utilization.

Traffic Engineering has been the prime concern for Internet Service Providers (ISPs), with the main focus being minimization of over-utilization of network capacity even though additional capacity is available which is under-utilized, Furthermore, requirements of timely delivery of digitized audiovisual information raises a new challenge of finding a path meeting these requirements. This paper addresses the issue of (a) distributing load to achieve global efficiency in resource utilization. (b) Finding a path satisfying the real time requirements of, delay and bandwidth requested by the applications. In this paper we do a critical study of the link utilization that varies over time and determine the time interval during which the link occupancy remains constant across days. This information helps in pre-determining link utilization that is useful in balancing load in the network Finally, we run simulations that use a dynamic time interval for profiling traffic and show improvement in terms number of calls admitted/blocked.

Maintaining dynamic sequences under equality tests in polylogarithmic time

We present a randomized and a deterministic data structure for maintaining a dynamic family of sequences under equality tests of pairs of sequences and creations of new sequences by joining or splitting existing sequences. Both data structures support equality tests in O(1) time. The randomized version supports new sequence creations in O(log(2) n) expected time where n is the length of the sequence created. The deterministic solution supports sequence creations in O(log n (log m log* m + log n)) time for the mth operation.

Dynamic path profile aided recompilation in a JAVA just-in-time compiler

Just-in-Time (JIT) compilers for Java can be augmented by making use of runtime profile information to produce better quality code and hence achieve higher performance. In a JIT compilation environment, the profile information obtained can be readily exploited in the same run to aid recompilation and optimization of frequently executed (hot) methods. This paper discusses a low overhead path profiling scheme for dynamically profiling AT produced native code. The profile information is used in recompilation during a subsequent invocation of the hot method. During recompilation tree regions along the hot paths are enlarged and instruction scheduling at the superblock level is performed. We have used the open source LaTTe AT compiler framework for our implementation. Our results on a SPARC platform for SPEC JVM98 benchmarks indicate that (i) there is a significant reduction in the number of tree regions along the hot paths, and (ii) profile aided recompilation in LaTTe achieves performance comparable to that of adaptive LaTTe in spite of retranslation and profiling overheads.

Molecular dynamic simulations of elastomer structure and its influence on anisotropic stress under time-varying strain

We investigate the evolution of polymer structure and its influence on uniaxial anisotropic stress under time-varying uniaxial strain, and the role of external control variables such as temperature, strain rate, chain length, and density, using molecular dynamics simulation. At temperatures higher than glass transition, stress anisotropy in the system is reduced even though the bond stretch is greater at higher temperatures. There is a significant increase in the stress level with increasing density. At higher densities, the uncoiling of the chains is suppressed and the major contribution to the deformation is by internal deformation of the chains. At faster rates of loading stress anisotropy increases. The deformation mechanism is mostly due to bond stretch and bond bending rather than overall shape and size. Stress levels increase with longer chain length. There is a critical value of the functionality of the cross-linkers beyond which the uniaxial stress developed increases caused primarily by bond stretching due to increased constraint on the motion of the monomers. Stacking of the chains in the system also plays a dominant role in the behaviour in terms of excluded volume interactions. Low density, high temperature, low values of functionality of cross-linkers, and short chain length facilitate chain uncoiling and chain slipping in cross-linked polymers.