Dynamic time warping

About: Dynamic time warping is a research topic. Over the lifetime, 6013 publications have been published within this topic receiving 133130 citations.

On design and implementation of an embedded automatic speech recognition system

Abstract: We present a new design of an Embedded Speech Recognition System. It combines the aspects of both hardware and software design to implement a speaker dependent, isolated word, small vocabulary speech recognition system. The feature extraction is based on modified Mel-scaled Frequency Cepstral Coefficients (MFCC) and template matching employs Dynamic Time Warping (DTW). A novel algorithm has been used to improve the detection of start of a word. The hardware is built around the industry standard TMS320LF2407A DSP. The board is designed to serve as a general purpose DSP development board for the 24X series of TI DSPs. It contains, apart from the DSP, the external SRAM, FLASH, ADC interface, I/O interfacing blocks and JTAG interface. Both the hardware and the software have been designed concurrently, with a view to achieve high-speed recognition with maximum accuracy in minimum power and making the device portable. The proposed solution is a low-cost, high-performance, scalable alternative to other existing products.

Heart sound classification based on temporal alignment techniques

Abstract: The ability to accurately stratify patients at risk of adverse cardiovascular outcomes using heart sound recordings could result in earlier treatment and improved patient outcomes. However, there remain several challenges associated with risk stratifying patients based on the phono-cardiogram (PCG) alone. First, inter-patient differences can make it challenging to learn a model that generalizes well across patients. Second, heterogeneity introduced by the collection environment of the recordings can render a classifier trained on one population useless when applied to another To address these challenges we explore the use of temporal alignment techniques, in particular dynamic time warping (DTW). Using DTW we compare heart sounds within and across subjects/recordings. These DTW based features, coupled with widely used spectral MFCC coefficients, serve as input to a linear SVM. Applied to the held-out test set our classifier obtained a test score of 82.4%, suggesting that temporal alignment techniques can effectively reduce the effects of inter-patient variability and mitigate the differences introduced by heterogeneous data collection environments.

On-line prediction of border crossing traffic using an enhanced Spinning Network method

Abstract: This paper improves on the Spinning Network (SPN) method, a novel forecasting technique, inspired by human memory which was recently developed by Huang and Sadek (2009) The improvement centers on the use of the Dynamic Time Warping (DTW) algorithm to assess the similarity between two given time series, instead of using the Euclidean Distance as was the case with the original SPN Following this, the enhanced method (ie, hereafter referred to as the DTW–SPN) is used to predict hourly traffic volumes at the Peace Bridge, an international border crossing connecting Western New York State in the US and Southern Ontario in Canada The performance of the DTW–SPN is then compared to that of three other forecasting methods, namely: (1) the original SPN (referred to as the Euclidean–SPN); (2) the Seasonal Autoregressive Integrated Moving Average (SARIMA) method; and (3) Support Vector Regression (SVR) Both classified as well as non-classified datasets are utilized, with the classification made on the basis of the type of the day to which the data items belong (ie Mondays through Thursdays, Fridays, weekends, holidays, and game days) The results indicate that, in terms of the Mean Absolute Percent Error, the DTW–SPN performed the best for all data groups with the exception of the “game day” group, where SVR performed slightly better From a computational efficiency standpoint, the SPN-type algorithms require runtime significantly lower than that for either SARIMA or SVR The performance of the DTW–SPN was also quite acceptable even when the data was not classified, indicating the robustness of the proposed forecasting method in dealing with heterogeneous data

An application-dependent framework for the recognition of high-level surgical tasks in the OR

Abstract: Surgical process analysis and modeling is a recent and important topic aiming at introducing a new generation of computer-assisted surgical systems Among all of the techniques already in use for extracting data from the Operating Room, the use of image videos allows automating the surgeons' assistance without altering the surgical routine We proposed in this paper an application-dependent framework able to automatically extract the phases of the surgery only by using microscope videos as input data and that can be adaptable to different surgical specialties First, four distinct types of classifiers based on image processing were implemented to extract visual cues from video frames Each of these classifiers was related to one kind of visual cue: visual cues recognizable through color were detected with a color histogram approach, for shape-oriented visual cues we trained a Haar classifier, for texture-oriented visual cues we used a bag-of-word approach with SIFT descriptors, and for all other visual cues we used a classical image classification approach including a feature extraction, selection, and a supervised classification The extraction of this semantic vector for each video frame then permitted to classify time series using either Hidden Markov Model or Dynamic Time Warping algorithms The framework was validated on cataract surgeries, obtaining accuracies of 95%

An EMG-based handwriting recognition through dynamic time warping

Abstract: In this paper, an electromyography (EMG)-based handwriting recognition method was proposed for a latent tendency of natural user interface. The subjects wrote the characters at a normal speed, and six channels of EMG signals were recorded from forearm muscles. The dynamic time warping (DTW) algorithm was used to eliminate the time axis variance during writing. The process for template making and matching was illustrated diagrammatically. The results showed that no more than ten training trials per character could make an accuracy of above 90%. The recognition performance was compared in three character sets: digits, Chinese characters and capital letters.