Sekharjit Datta

Mel filter-like admissible wavelet packet structure for speech recognition

A new filter structure using admissible wavelet packet analysis is presented. These filters have the advantage of having frequency bands spacing similar to the Mel scale. Further wavelet packet transforms multiresolution capabilities are used to derive new sets of features, which are found to be superior to Mel frequency cepstral coefficients (MFCC) in unvoiced phoneme classification problems.

Phoneme recognition using wavelet based features

This paper proposes the use of the discrete wavelet transform (DWT) for the extraction of features from phonemes. Instead of using the short time Fourier transform for feature extraction a new set of features is obtained from the DWT. The new set of features overcomes the previously reported problem of shift variance in DWT based features. Training and test samples of the phonemes were obtained from the TIMIT database. To account for the fast changes in the phonemes, the features were calculated for different phoneme durations and the performance was compared. For the classification of the phonemes, two different classifiers were used, based on linear discriminant analysis and multi-layer perceptron.

WAVELET SUB-BAND BASED TEMPORAL FEATURES FOR ROBUST HINDI PHONEME RECOGNITION

This paper proposes the use of wavelet transform-based feature extraction technique for Hindi speech recognition application. The new proposed features take into account temporal as well as frequency band energy variations for the task of Hindi phoneme recognition. The recognition performance achieved by the proposed features is compared with the standard MFCC and 24-band admissible wavelet packet-based features using a linear discriminant function based classifier. To evaluate robustness of these features, the NOISEX database is used to add different types of noise into phonemes to achieve signal-to-noise ratios in the range of 20 dB to -5 dB. The recognition results show that under noisy background the proposed technique always achieves a better performance over MFCC-based features.

Monitoring heart disease and diabetes with mobile internet communications

A telemedicine system is described for monitoring vital signs and general health indicators of patients with cardiac and diabetic conditions. Telemetry from wireless sensors and readings from other instruments are combined into a comprehensive set of measured patient parameters. Using a combination of mobile device applications and web browser, the data can be stored, accessed, and displayed using mobile internet communications to the central server. As an extra layer of security in the data transmission, information embedded in the data is used in its verification. The paper highlights features that could be enhanced from previous systems by using alternative components or methods.

Data acquisition in a wireless diabetic and cardiac monitoring system

A telemedicine system is described for monitoring the vital signs and general health indicators of patients with cardiac and diabetic conditions. Telemetry from wireless sensors and readings from other instruments are combined into a comprehensive patient health dataset. The data can be stored, accessed and displayed using mobile Internet communications with a server. The paper concentrates on the data acquisition process, using an alternative sensor network protocol to Bluetooth and manual data entry into a smartphone application and HTML5 web browser.

Wrist pulse signal classification for health diagnosis

Ancient Indian and Chinese medicine both use non-invasive wrist pulse signals for health diagnosis of patients. In this paper, data obtained from a number of patients have been used to categorize the types of pulse signals that are found in both normal and abnormal health conditions. Features were extracted from the pulse signals using both frequency and wavelet transformations and these were then ranked according to their classification power for multiclass classifier design. Linear and quadratic pulse classifiers are proposed with raw features as well as subset of ranked features. Linear classifier has found to be giving highest classification accuracy of 73.82% using 4 ranked features.

Development of m-health monitoring systems in India and Iraq

Two separate projects have been carried out to implement m-health programs in India and Iraq, and, for each, this paper describes the work performed by the teams involved, presents results and details a number of lessons learned. In general, it is found that although India and Iraq have very different medical priorities, they pose similar issues when introducing m-health strategies.

Mobile communications for monitoring heart disease and diabetes

This paper describes a practical development project to enable the monitoring of vital signs data obtained from patients located in remote rural locations. The data are gathered from a wireless network of sensors attached to a patients body and stored locally for secure transmission over existing communication infrastructures to a hospital server. Clinicians are then able to monitor the patient offline and upload diagnoses.

Wavelet Based Sub-space Features for Face Recognition

In this paper we propose features based on sub-space projection methods using Principal Component Analysis (PCA) and Independent Component Analysis (ICA) on wavelet sub-band for face recognition. Wavelet based sub-band decomposition helps to reduce the size of image, and the approximate image obtained in the low-low (approximate) band is used here to apply sub-space projection methods. This improves the speed of feature extraction process without compromising the recognition performance. Classification of the faces based on the extracted features was carried out by using a Linear Discriminant function based classifier on Olivetti Research Laboratory (ORL) image database. Different level of wavelet decomposition is carried out and recognition performance evaluated. Highest recognition was achieved at 3 level wavelet decomposition using ICA. The proposed scheme uses minimum number of features and the recognition results obtained show an improvement of about 0.5% over some of the existing schemes with lower computation cost.

A comparison of visual features for audiovisual automatic speech recognition

The use of visual information from speakers mouth region have shown to improve the performance of automatic speech recognition (ASR) systems. This is particularly important in presence of noise which even in moderate form severely degrades the speech recognition performance of systems using only audio information. Various sets of features extracted from speakers mouth region have been used to improve upon the performance of an ASR system in such challenging conditions and have met many successes. To the best of authors knowledge, the effect of using these techniques on recognition performance on the basis of phonemes have not been investigated yet. This paper presents a comparison of phoneme recognition performance using visual features extracted from mouth region‐of‐interest using discrete cosine transform (DCT) and discrete wavelet transform (DWT). New DCT and DWT features have also been extracted and compared with the previously used one. These features were used along with audio features based on Me...