An Efficient and Perceptually Motivated Auditory Neural Encoding and Decoding Algorithm for Spiking Neural Networks.

The auditory front-end is an integral part of a spiking neural network (SNN) when performing auditory cognitive tasks. It encodes the temporal dynamic stimulus, such as speech and audio, into an efficient, effective and reconstructable spike pattern to facilitate the subsequent processing. However, most of the auditory front-ends in current studies have not made use of recent findings in psychoacoustics and physiology concerning human listening. In this paper, we propose a neural encoding and decoding scheme that is optimized for audio processing. The neural encoding scheme, that we call Biologically plausible Auditory Encoding (BAE), emulates the functions of the perceptual components of the human auditory system, that include the cochlear filter bank, the inner hair cells, auditory masking effects from psychoacoustic models, and the spike neural encoding by the auditory nerve. We evaluate the perceptual quality of the BAE scheme using PESQ; the performance of the BAE based on sound classification and speech recognition experiments. Finally, we also built and published two spike-version of speech datasets: the Spike-TIDIGITS and the Spike-TIMIT, for researchers to use and benchmarking of future SNN research.

Blood Pressure Estimation Using Time Domain Features of Auscultatory Waveforms and GMM-HMM Classification Approach.

This paper presents a novel method to estimate systolic blood pressure (SBP) and diastolic blood pressure (DBP) from time domain features extracted from auscultatory waveforms (AWs) and using a Gaussian Mixture Models and Hidden Markov Model (GMM-HMM) classification approach. The three time domain features selected include the cuff pressure (CP), the energy of the Korotkoff pulses (KE), and the slope of the KE (SKE). The proposed GMM-HMM can effectively discover the latent structure in AW sequences and automatically learn such structures. The SBP and DBP points are then detected as the cuff pressures at which AW sequence changes its structure. We conclude that the proposed GMM-HMM estimation method is a very promising method improving the accuracy of automated non-invasive measurement of blood pressure.

Novel methods of testing and calibration of oscillometric blood pressure monitors.

We present a robust method for testing and calibrating the performance of oscillometric non-invasive blood pressure (NIBP) monitors, using an industry standard NIBP simulator to determine the characteristic ratios used, and to explore differences between different devices. Assuming that classical auscultatory sphygmomanometry provides the best approximation to intra-arterial pressure, the results obtained from oscillometric measurements for a range of characteristic ratios are compared against a modified auscultatory method to determine an optimum characteristic ratio, Rs for systolic blood pressure (SBP), which was found to be 0.565. We demonstrate that whilst three Chinese manufactured NIBP monitors we tested used the conventional maximum amplitude algorithm (MAA) with characteristic ratios Rs = 0.4624±0.0303 (Mean±SD) and Rd = 0.6275±0.0222, another three devices manufactured in Germany and Japan either do not implement this standard protocol or used different characteristic ratios. Using a reference database of 304 records from 102 patients, containing both the Korotkoff sounds and the oscillometric waveforms, we showed that none of the devices tested used the optimal value of 0.565 for the characteristic ratio Rs, and as a result, three of the devices tested would underestimate systolic pressure by an average of 4.8mmHg, and three would overestimate the systolic pressure by an average of 6.2 mmHg.

Improving the quality and accuracy of non-invasive blood pressure measurement by visual inspection and automated signal processing of the Korotkoff sounds.

OBJECTIVE: In this study we investigate inter-operator differences in determining systolic and diastolic pressure from auscultatory sound recordings of Korotkoff sounds. We introduce a new method to record and convert Korotkoff sounds to a high fidelity sound file which can be replayed under optimal conditions by multiple operators, for the independent determination of systolic and diastolic pressure points. APPROACH: We have developed a digitised data base of 643 NIBP records from 216 subjects. The Korotkoff signals of 310 good quality records were digitised and the Korotkoff sounds converted to high fidelity audio files. A randomly selected subset of 90 of these data files, were used by an expert panel to independently detect systolic and diastolic points. We then developed a semi-automated method of visualising processed Korotkoff sounds, supported by simple algorithms to detect systolic and diastolic pressure points that provided new insights on the reasons for large differences recorded by the expert panel. MAIN RESULTS: Detailed analysis of the 90 randomly selected records revealed that peak root mean square (RMS) energy of the Korotkoff sounds, ranged from 3.3 to 84 mV rms, with the lower bound below the audible range of 4-6 mV rms. The diastolic phase was below the minimum auditory threshold in only 47/90 records. This indicates that for approximately 50% of all records diastole could not be determined from Phase V silence. The maximum relative error recorded for systolic pressure between the two methods, auscultatory and visual/algorithmic, was 30.8 mmHg with a mean error of 8.0 ± 5.4 mmHg. We explore the impact of signal morphology and intensity of the Korotkoff sounds, as well as noise, cardiac arrhythmia and the hearing acuity of the operator, on the accuracy of the measurement. SIGNIFICANCE: We conclude that large intra-personal variability in Korotkoff signal morphology and amplitudes, as well as variations in the hearing acuity of the operator, make accurate NIBP measurements using sphygmomanometry difficult and should not be used as the gold standard against which automated NIBP devices are calibrated. We propose an alternative method of visualizing the energy of the Korotkoff sounds and applying simple algorithms to determine systolic and diastolic pressure points, which whilst mimicking classical sphygmomanometry eliminates the problems associated with operator hearing acuity and complex and variable Korotkoff signal morphology.

Non-Invasive measurement of blood pressure - Why we should look at BP traces rather than listen to Korotkoff sounds.

Accurate non-invasive measurement of blood pressure in unsupervised environments continues to be a challenge, particularly in the presence of movement artefact, electrical noise and most importantly cardiac arrhythmia which are common in those aged over 65 suffering from a range of chronic conditions. Large intra personal variability in signal morphometry and amplitudes further complicates the development of reliable signal processing algorithms for NIBP measurement. In this paper we demonstrate the effect of this variability and propose that the traditional methods of human blood pressure determination by sphygmomanometry should no longer be considered a gold standard for the calibration of NIBP devices.

The effect of automatic gain control structure and release time on cochlear implant speech intelligibility.

Nucleus cochlear implant systems incorporate a fast-acting front-end automatic gain control (AGC), sometimes called a compression limiter. The objective of the present study was to determine the effect of replacing the front-end compression limiter with a newly proposed envelope profile limiter. A secondary objective was to investigate the effect of AGC speed on cochlear implant speech intelligibility. The envelope profile limiter was located after the filter bank and reduced the gain when the largest of the filter bank envelopes exceeded the compression threshold. The compression threshold was set equal to the saturation level of the loudness growth function (i.e. the envelope level that mapped to the maximum comfortable current level), ensuring that no envelope clipping occurred. To preserve the spectral profile, the same gain was applied to all channels. Experiment 1 compared sentence recognition with the front-end limiter and with the envelope profile limiter, each with two release times (75 and 625 ms). Six implant recipients were tested in quiet and in four-talker babble noise, at a high presentation level of 89 dB SPL. Overall, release time had a larger effect than the AGC type. With both AGC types, speech intelligibility was lower for the 75 ms release time than for the 625 ms release time. With the shorter release time, the envelope profile limiter provided higher group mean scores than the front-end limiter in quiet, but there was no significant difference in noise. Experiment 2 measured sentence recognition in noise as a function of presentation level, from 55 to 89 dB SPL. The envelope profile limiter with 625 ms release time yielded better scores than the front-end limiter with 75 ms release time. A take-home study showed no clear pattern of preferences. It is concluded that the envelope profile limiter is a feasible alternative to a front-end compression limiter.

Robust sounds of activities of daily living classification in two-channel audio-based telemonitoring.

Despite recent advances in the area of home telemonitoring, the challenge of automatically detecting the sound signatures of activities of daily living of an elderly patient using nonintrusive and reliable methods remains. This paper investigates the classification of eight typical sounds of daily life from arbitrarily positioned two-microphone sensors under realistic noisy conditions. In particular, the role of several source separation and sound activity detection methods is considered. Evaluations on a new four-microphone database collected under four realistic noise conditions reveal that effective sound activity detection can produce significant gains in classification accuracy and that further gains can be made using source separation methods based on independent component analysis. Encouragingly, the results show that recognition accuracies in the range 70%-100% can be consistently obtained using different microphone-pair positions, under all but the most severe noise conditions.

Novel delta zero crossing regression features for gait pattern classification.

Many recent research works on gait pattern classification indicates that static features are used. This paper describes of extracting novel dynamic features as complimentary features for the gait pattern classification. The dynamic features are obtained by using regression on the delta zero crossing counts (ΔZCC) of the acceleration signal. The classification results using the filterbank features with the novel dynamic features showed an overall accuracy of 97% was achieved. This is an improvement of 3% from using the filterbank features alone.

Can triaxial accelerometry accurately recognize inclined walking terrains?

The standard method for the analysis of body accelerations cannot accurately estimate the energy expenditure (EE) of uphill or downhill walking. The ability to recognize the grade of the walking surface will most likely improve upon the accuracy of the EE estimates for daily physical activities. This paper investigates the benefits of automatic gait analysis approaches including step-by-step gait segmentation and heel-strike recognition of the accelerometry signal in classifying various gradients. Triaxial accelerometry signals were collected from 12 subjects, performing walking on seven different gradient surfaces: 1) 92 m of 0(°) flat ground; 2) 85 m of ±2.70(°) inclined ramp; 3) 24 m of ±9.86(°) inclined ramp; and 4) 6-m pitch line of ±28.03(°) rake of stairway. Validity studies performed on a group of randomly selected healthy subjects showed high agreement scores between the automated heel-strike recognition markers, manual gait annotation markers, and video-based gait-segmentation markers. Thirteen subset features were found using a subset-selection search procedure from 57 extracted features which maximize the classification accuracy, performed with a Gaussian mixture model classifier, as estimated using sixfold cross-validation. An overall walking pattern-recognition accuracy of 82.46% was achieved on seven different inclined terrains using the 13 selected features. This system should, therefore, improve the accuracy of daily EE estimates with accurate measures on terrain inclinations.

Gait pattern classification using compact features extracted from intrinsic mode functions.

Recent research work indicates that gait patterns are both non-linear and non-stationary signals and they can be analyzed using empirical mode decomposition. This paper describes gait pattern classification using features that are obtained by performing discrete cosine transforms (DCT) on intrinsic mode functions of five different human gait patterns. The DCT provides a compact 8-dimensional feature vector for gait pattern classification. Fifty two subjects participated in the experiment. The classification was performed using a Gaussian mixture model and an overall accuracy of 90.2% was achieved.

Dual-microphone Sounds of Daily Life classification for telemonitoring in a noisy environment.

Telemonitoring of elderly people in their homes using video cameras is complicated by privacy concerns, and hence sound has emerged as a promising alternative that is more acceptable to users. We investigate methods to address the accuracy degradation of sound classification that arises in the presence of background noise typical of a practical telemonitoring situation. A dual microphone configuration is used to record a database of Sounds of Daily Life (SDL) in a kitchen. We introduce a new algorithm employing the eigenvalues of the cross-spectral matrix of the recorded signals to detect the endpoints of a SDL in the presence of background noise. Independent component analysis is also used to improve the signal to noise ratio of the SDL. Results on a 7-class noisy SDL classification problem show that the error rate the proposed SDL classification system can be improved by up to 97% relative to a single-microphone system without noise reduction techniques, when evaluated on a large SDL database with SNRs in the range 0-28 dB.

Accelerometry based classification of walking patterns using time-frequency analysis.

In this work, 33 dimensional time-frequency domain features were developed and evaluated to detect five different human walking patterns from data acquired using a triaxial accelerometer attached at the waist above the iliac spine. 52 subjects were asked to walk on a flat surface along a corridor, walk up and down a flight of a stairway and walk up and down a constant gradient slope, in an unsupervised manner. Time-frequency domain features of acceleration data in anterior-posterior (AP), medio-lateral (ML) and vertical (VT) direction were developed. The acceleration signal in each direction was decomposed to six detailed signals at different wavelet scales by using the wavelet packet transform. The rms values and standard deviations of the decomposed signals at scales 5 to 2 corresponding to the 0.78-18.75 Hz frequency band were calculated. The energies in the 0.39-18.75 Hz frequency band of acceleration signal in AP, ML and VT directions were also computed. The back-end of the system was a multi-layer perceptron (MLP) Neural Networks (NNs) classifier. Overall classification accuracies of 88.54% and 92.05% were achieved by using a round robin (RR) and random frame selecting (RFS) train-test method respectively for the five walking patterns.

Classification of a known sequence of motions and postures from accelerometry data using adapted Gaussian mixture models.

Accelerometry shows promise in providing an inexpensive but effective means of long-term ambulatory monitoring of elderly patients. The accurate classification of everyday movements should allow such a monitoring system to exhibit greater 'intelligence', improving its ability to detect and predict falls by forming a more specific picture of the activities of a person and thereby allowing more accurate tracking of the health parameters associated with those activities. With this in mind, this study aims to develop more robust and effective methods for the classification of postures and motions from data obtained using a single, waist-mounted, triaxial accelerometer; in particular, aiming to improve the flexibility and generality of the monitoring system, making it better able to detect and identify short-duration movements and more adaptable to a specific person or device. Two movement classification methods were investigated: a rule-based Heuristic system and a Gaussian mixture model (GMM)-based system. A novel time-domain feature extraction method is proposed for the GMM system to allow better detection of short-duration movements. A method for adapting the GMMs to compensate for the problem of limited user-specific training data is also proposed and investigated. Classification performance was considered in relation to data gathered in an unsupervised, directed routine conducted in a three-month field trial involving six elderly subjects. The GMM system was found to achieve a mean accuracy of 91.3%, distinguishing between three postures (sitting, standing and lying) and five movements (sit-to-stand, stand-to-sit, lie-to-stand, stand-to-lie and walking), compared to 71.1% achieved by the Heuristic system. The adaptation method was found to offer a mean accuracy of 92.2%; a relative improvement of 20.2% over tests without subject-specific data and 4.5% over tests using only a limited amount of subject-specific data. While limited to a restricted subset of possible motions and postures, these results provide a significant step in the search for a more robust and accurate ambulatory classification system.

An adapted Gaussian mixture model approach to accelerometry-based movement classification using time-domain features.

The accurate classification of everyday movements from accelerometry data will provide a significant step towards the development of effective ambulatory monitoring systems for falls detection and prediction. The search continues for optimal front-end processing methods for use in accelerometry systems. Here, we propose a novel set of time domain features, which achieve a mean accuracy of 91.3% in distinguishing between three postures (sitting, standing and lying) and five movements (sit-to-stand, stand-to-sit, lie-to-stand, stand-to-lie and walking). This is a 39.2% relative improvement in error rate over more commonly used frequency based features. A method for adapting Gaussian Mixture Models to compensate for the problem of limited user-specific training data is also proposed and investigated. The method, which uses Bayesian adaptation, was found to improve classification performance for time domain features, offering a mean relative improvement of 20.2% over a non subject-specific system and 4.5% over a system trained using subject specific data only.