Sourabh Ravindran

Motion artifact cancellation to obtain heart sounds from a single chest-worn accelerometer

This paper presents a method of extracting primary heart sound signals from chest-worn accelerometer data in the presence of motion artifacts. The proposed method outperforms noise removal techniques such as wavelet denoising and adaptive filtering. Results from six subjects show a primary heart signal detection rate of 99.36% with a false positive rate of 1.3%.

Gain adaptation based on signal-to-noise ratio for noise suppression

In this paper we describe a technique that uses adaptive gain control to achieve noise suppression in speech signals. The method used to map the dynamic range of the signal is based on the human auditory perceptual model. Since the processing is based on the model of human perception, the resulting noise suppressed speech is natural sounding. The computational complexity of the proposed method is low and the mapping of the dynamic range of the signal has a low delay. Because of these properties, this method is ideal for real-time implementation.

Low complexity algorithms for heart rate and epileptic seizure detection

Algorithm design for low power platforms is constrained by memory and computational limitations, and real-world applications demand robust performance. This paper presents two algorithms that were designed with the view that simplicity can translate to robustness. The first algorithm processes electrocardiogram (ECG) signals to detect QRS complexes reliably in the presence of significant noise. The second algorithm is a low-cost approach to detecting seizure onset from electro-corticogram (ECoG) data. The ECG algorithm was implemented on a TI MSP430-based platform and the ECoG algorithm was implemented (in simulation) on a Cortex M3 based ultra-low power device.

Real-time implementation of a chest-worn accelerometer based heart monitoring system

This paper presents the real-time implementation of an approach for monitoring the heart using a single accelerometer. The implementation described is based on an algorithmic approach that has been shown to successfully remove motion artifacts from accelerometer based heart signal measurements. A driving factor in this project is the need for low-cost heart rate monitoring as part of a personal monitoring system for use world-wide. The digital signal processor (DSP) chosen for the implementation was Texas Instruments (TI) TMS320C5505 DSP in order to reduce the power requirement of the specific implementation thereby conserving battery life. The TMS320C5505 CPU can operate on 1.05V and the chip architecture has been highly optimized for energy efficiency.

Chest-accelerometry for hemodynamic trending during valsalva-recovery

Chest-worn accelerometers have been shown to detect acoustic and mechanical signals corresponding to cardiovascular activity. This paper aims at investigating and characterizing two different components of chest acceleration (seismocardiogram) along two orthogonal axes: firstly, the sub-10 Hz ballistic signal components dominant in the vertical axis and secondly, the 10–50 Hz acoustic signal components more dominantly expressed in the radial axis. Acceleration signals from five subjects in response to a valsalva maneuver were measured. Correlations of features from the two above acceleration components were computed with respect to reference measurements of stroke volume and pulse pressure obtained with a Finapres continuous blood pressure system. The peak amplitude of the vertical ballistic and radial acoustic signal components were found to correlate well with stroke volume (R=0.78 and 0.83, for vertical ballistic and radial acoustic, respectively). Comparable correlations were found between beat RMS power (R=0.77 and 0.83) and stroke volume. Similarly, correlations were also observed between pulse pressure and peak amplitude (R=0.74 and 0.86) and the beat RMS power (R=0.74 and 0.86).

Real-time, low-complexity, low-memory solution to ECG-based heart rate detection

This paper addresses the issue of heart rate detection from noisy ECG data, and presents a method with low complexity and low memory requirements that can detect QRS complex in the presence of noise and muscle artifacts. On the MIT-BIH arrhythmia database we were able to detect 99.3% of QRS complexes with 0.47% false detection. This method can also be applied to heart rate detection using phonocardio signals.

Heart rate and epileptic seizure detection algorithms for low-power platforms

Abstract Algorithm design for low power platforms is constrained by memory and computational limitations, and real-world applications demand robust performance. This paper presents two algorithms that were designed with the view that simplicity can translate to robustness. The first algorithm processes electrocardiogram (ECG) signals to detect QRS complexes reliably in the presence of significant noise. The second algorithm is a low-cost approach to detecting seizure onset from electrocorticogram (ECoG) data. The ECG algorithm was implemented on a TI MSP430-based platform and the ECoG algorithm was implemented on a Cortex-M3 based ultra-low power device.