Akinori Nishihara

A framework for automatic heart sound analysis without segmentation

BackgroundA new framework for heart sound analysis is proposed. One of the most difficult processes in heart sound analysis is segmentation, due to interference form murmurs.MethodEqual number of cardiac cycles were extracted from heart sounds with different heart rates using information from envelopes of autocorrelation functions without the need to label individual fundamental heart sounds (FHS). The complete method consists of envelope detection, calculation of cardiac cycle lengths using auto-correlation of envelope signals, features extraction using discrete wavelet transform, principal component analysis, and classification using neural network bagging predictors.ResultThe proposed method was tested on a set of heart sounds obtained from several on-line databases and recorded with an electronic stethoscope. Geometric mean was used as performance index. Average classification performance using ten-fold cross-validation was 0.92 for noise free case, 0.90 under white noise with 10 dB signal-to-noise ratio (SNR), and 0.90 under impulse noise up to 0.3 s duration.ConclusionThe proposed method showed promising results and high noise robustness to a wide range of heart sounds. However, more tests are needed to address any bias that may have been introduced by different sources of heart sounds in the current training set, and to concretely validate the method. Further work include building a new training set recorded from actual patients, then further evaluate the method based on this new training set.

Universal maximally flat lowpass FIR systems

The family of FIR digital filters with maximally flat magnitude and group delay response is considered. The filters were proposed by Baher (1982), who furnished them with an analytic procedure for derivation of their transfer function. The contributions of this paper are the following. A simplified formula is presented for the transfer function of the filters. The equivalence of the novel formula with a formula that is derived from Bahers analytical procedure is proved using a modern method for automatic proof of identities involving binomial coefficients. The universality of Bahers filters is then established by proving that they include linear-phase filters, generalized half-band filters, and fractional delay systems. In this way, several classes of maximally flat filters are unified under a single formula. The generating function of the filters is also derived. This enables us to develop multiplierless cellular array structures for exact realization of a subset of the filters. The subset that enjoys such multiplierless realizations includes linear-phase filters, some nonsymmetric filters, and generalized halfband filters. A procedure for designing the cellular array structures is also presented.

Efficient voice activity detection algorithm using long-term spectral flatness measure

This paper proposes a novel and robust voice activity detection (VAD) algorithm utilizing long-term spectral flatness measure (LSFM) which is capable of working at 10 dB and lower signal-to-noise ratios(SNRs). This new LSFM-based VAD improves speech detection robustness in various noisy environments by employing a low-variance spectrum estimate and an adaptive threshold. The discriminative power of the new LSFM feature is shown by conducting an analysis of the speech/non-speech LSFM distributions. The proposed algorithm was evaluated under 12 types of noises (11 from NOISEX-92 and speech-shaped noise) and five types of SNR in core TIMIT test corpus. Comparisons with three modern standardized algorithms (ETSI adaptive multi-rate (AMR) options AMR1 and AMR2 and ITU-T G.729) demonstrate that our proposed LSFM-based VAD scheme achieved the best average accuracy rate. A long-term signal variability (LTSV)-based VAD scheme is also compared with our proposed method. The results show that our proposed algorithm outperforms the LTSV-based VAD scheme for most of the noises considered including difficult noises like machine gun noise and speech babble noise.

Maximally flat half-band diamond-shaped FIR filters using the Bernstein polynomial

Two-dimensional (2-D) linear-phase diamond-shaped (DS) FIR digital filters have important image processing applications. They are used in sampling structure conversion, e.g. interlace-to-progressive conversion. Maximally flat (MF) DS filters have attractive time- and frequency-domain properties and are preferred in some applications. In this correspondence MF half-band DS FIR are designed using the two-variable Bernstein polynomial. Formulas for the transfer function and the impulse response are obtained. >

A high-speed FIR digital filter with CSD coefficients implemented on FPGA

A very fast and low-complexity FIR digital filter on FPGA is presented. Multipliers in the filter whose coefficients are expressed as canonic signed digit (CSD) code are realized with wired-shifters, adders and subtracters. The critical path is minimized by insertion of pipeline registers and is equal to the propagation delay of an adder. The number of pipeline registers is limited by using an equivalent transformation on a signal flow graph. The price paid for the 100% speedup is 5% increase in the area. The maximum sampling frequency is 78.6 MHz.

Regular orthonormal and biorthogonal wavelet filters

Abstract New methods for half-band filter design are developed, which structurally incorporate the regularity constraint into the design procedure. The first method results in half-band filters which do not have strictly positive frequency response. It is suitable for one class of biorthogonal filter banks. The second method results in half-band filters with strictly positive frequency response from which orthonormal wavelet filters can be obtained by spectral factorization. All filters are regular and have sharp transition bands. A new factorization of the first polyphase component of every half-band filter is found. This factorization suggests an efficient implementation of biorthogonal filter banks of the type considered. The implementation not only preserves the perfect-reconstruction property assuming finite word-length arithmetic, but has reduced coefficient sensitivity compared with the direct form. The properties of the resulting scaling functions and wavelets are summarized.

Online secondary path modeling in multichannel active noise control systems using variable step size

In single-channel feedforward active noise control (ANC) systems, additive random noise based methods are often applied to achieve secondary path modeling (SPM) during online operation. This paper investigates the issue of online SPM in multichannel ANC systems. It is shown that the application of existing methods for online SPM in multichannel ANC systems greatly increases the computational complexity. Here we extend our previous work on single-channel variable step-size online SPM to multichannel ANC systems. It is shown that the proposed method has reduced computational complexity as compared with other methods. Computer simulations are carried out that demonstrate the effectiveness of the proposed method.

Maximally flat FIR filters

A new Bernstein approximant for minimax filters is suggested. Explicit solutions for maximally flat and monotonic filters are obtained as a special case of the new Bernstein approximant for all linear-phase finite impulse response (FIR) filter types. A design technique starting from arbitrary specifications is described. Using the McClellan transformation, multidimensional filters are obtained.<<ETX>>

Explicit Formula for Predictive FIR Filters and Differentiators Using Hahn Orthogonal Polynomials

An explicit expression for the impulse response coefficients of the predictive FIR digital filters is derived. The formula specifies a four-parameter family of smoothing FIR digital filters containing the Savitsky-Goaly filters, the Heinonen-Neuvo polynomial predictors, and the smoothing differentiators of arbitrary integer orders. The Hahn polynomials, which are orthogonal with respect to a discrete variable, are the main tool employed in the derivation of the formula. A recursive formula for the computation of the transfer function of the filters, which is the z-transform of a terminated sequence of polynomial ordinates, is also introduced. The formula can be used to design structures with low computational complexity for filters of any order.

Multiplierless and hierarchical structures for maximally flat half-band FIR filters

A simple method to derive a closed-form expression for the transfer function of linear-phase half-band filters with maximally flat amplitude-response characteristics is presented. The method is based on the binomial series. It results in hierarchical and modular structures with low hardware complexity for low-pass and high-pass filters. For a filter of a given order, the structures provide access to ail maximally flat filters of lower orders. Two types of structures are presented. The first enjoys a set of multiplier coefficients with reduced dynamic range, and the second can be realized free of multiplier coefficients in a modular manner. Extension of the order of the filter can be achieved by cascading additional building blocks for the former structure or by adding an extra layer of modules for the latter structure. The proposed structures and formulas are applicable to maximally flat Hilbert transformers as well.