Maria Sandsten

Time-frequency image enhancement based on interference suppression in Wigner-Ville distribution

This paper proposes a time-frequency (t-f) image enhancement method for suppressing interference terms in the Wigner-Ville distribution. The proposed technique adapts the direction of the smoothing kernel locally at each t-f point, so that the smoothing kernel remains aligned with the ridges of the auto-terms. This local alignment of the smoothing kernel reduces cross-terms without degrading the energy concentration of auto-terms. The results indicate that the proposed time-frequency distribution outperforms other methods in terms of its ability to resolve close signal components. A new high resolution adaptive time-frequency distribution (TFD) is proposed.The proposed TFD adapts the direction of smoothing kernel on point by point basis.The proposed TFD outperforms other methods in terms of its ability to resolve close components.

Estimating time-varying RSA to examine psychophysiological linkage of marital dyads

One of the primary tenets of polyvagal theory dictates that parasympathetic influence on heart rate, often estimated by respiratory sinus arrhythmia (RSA), shifts rapidly in response to changing environmental demands. The current standard analytic approach of aggregating RSA estimates across time to arrive at one value fails to capture this dynamic property within individuals. By utilizing recent methodological developments that enable precise RSA estimates at smaller time intervals, we demonstrate the utility of computing time-varying RSA for assessing psychophysiological linkage (or synchrony) in husband-wife dyads using time-locked data collected in a naturalistic setting.

Automated analysis of song structure in complex birdsongs

Understanding communication and signalling has long been strived for in studies of animal behaviour. Many songbirds have a variable and complex song, closely connected to territory defence and reproductive success. However, the quantification of such variable song is challenging. In this paper, we present a novel, automated method for detection and classification of syllables in birdsong. The method provides a tool for pairwise comparison of syllables with the aim of grouping them in terms of their similarity. This allows analyses such as (1) determining repertoire size within an individual, (2) comparing song similarity between individuals within as well as between populations of a species and (3) comparing songs of different species (e.g. for species recognition). Our method is based on a particular feature representation of song units (syllables) which ensures invariance to shifts in time, frequency and amplitude. Using a single song from a great reed warbler, Acrocephalus arundinaceus, recorded in the wild, the proposed algorithm is evaluated by means of comparison to manual auditory and visual (spectrogram) song investigation by a human expert and to standard song analysis methods. Our birdsong analysis approach conforms well to manual classification and, moreover, outperforms the hitherto widely used methods based on mel-frequency cepstral coefficients and spectrogram cross-correlation. Thus, our algorithm is a methodological step forward for analyses of song (syllable) repertoires of birds singing with high complexity.

The Scaled Reassigned Spectrogram with Perfect Localization for Estimation of Gaussian Functions

The reassignment technique is used to increase localization for signal components in the time-frequency representation. The technique gives perfect localization for infinite linear chirp-signals, impulses and constant frequency signals but not for short non-stationary signals. In this paper, a scaled reassignment is proposed, based on the spectrogram using a Gaussian window. The resulting reassignment gives perfect localization for a Gaussian function when the window length matches the function length. Based on the scaled reassignment, an algorithm that estimates the Gaussian function length is also proposed.

Classification of one-dimensional non-stationary signals using the Wigner-Ville distribution in convolutional neural networks

In this paper we argue that the Wigner-Ville distribution (WVD), instead of the spectrogram, should be used as basic input into convolutional neural network (CNN) based classification schemes. The WVD has superior resolution and localization as compared to other time-frequency representations. We present a method where a large-size kernel may be learned from the data, to enhance features important for classification. We back up our claims with theory, as well as application on simulated examples and show superior performance as compared to the commonly used spectrogram.

Optimal time-frequency distributions using a novel signal adaptive method for automatic component detection

Finding objective methods for assessing the performance of time-frequency distributions (TFD) of measured multi-component signals is not trivial. An optimal TFD should have well resolved signal components (auto-terms) and well suppressed cross-terms. This paper presents a novel signal adaptive method, which is shown to have better performance than the existing method, of automatically detecting the signal components for TFD time instants of two-component signals. The method can be used together with a performance measure to receive automatic and objective performance measures for different TFDs, which allows for an optimal TFD to be obtained. The new method is especially useful for signals including auto-terms of unequal amplitudes and non-linear frequency modulation. The method is evaluated and compared to the existing method, for finding the optimal parameters of the modified B-distribution. The performance is also shown for an example set of Heart Rate Variability (HRV) signals. HighlightsA novel method to distinguish auto-terms and cross-terms in time-frequency analysis.Improves performance measures for objective comparison of time-frequency methods.Reliable also for auto-terms of unequal amplitudes and non-linear frequencies.Automatic optimal time-frequency analysis of HRV signals can be obtained.

Robust feature representation for classification of bird song syllables

A novel feature set for low-dimensional signal representation, designed for classification or clustering of non-stationary signals with complex variation in time and frequency, is presented. The feature representation of a signal is given by the first left and right singular vectors of its ambiguity spectrum matrix. If the ambiguity matrix is of low rank, most signal information in time direction is captured by the first right singular vector while the signal’s key frequency information is encoded by the first left singular vector. The resemblance of two signals is investigated by means of a suitable similarity assessment of the signals’ respective singular vector pair. Application of multitapers for the calculation of the ambiguity spectrum gives an increased robustness to jitter and background noise and a consequent improvement in performance, as compared to estimation based on the ordinary single Hanning window spectrogram. The suggested feature-based signal compression is applied to a syllable-based analysis of a song from the bird species Great Reed Warbler and evaluated by comparison to manual auditive and/or visual signal classification. The results show that the proposed approach outperforms well-known approaches based on mel-frequency cepstral coefficients and spectrogram cross-correlation.

Evaluation of Seven Time-Frequency Representation Algorithms Applied to Broadband Echolocation Signals

Time-frequency representation algorithms such as spectrograms have proven to be useful tools in marine biosonar signal analysis. Although there are several different time-frequency representation algorithms designed for different types of signals with various characteristics, it is unclear which algorithms that are best suited for transient signals, like the echolocation signals of echolocating whales. This paper describes a comparison of seven different time-frequency representation algorithms with respect to their usefulness when it comes to marine biosonar signals. It also provides the answer to how close in time and frequency two transients can be while remaining distinguishable as two separate signals in time-frequency representations. This is, for instance, relevant in studies where echolocation signal component azimuths are compared in the search for the exact location of their acoustic sources. The smallest time difference was found to be 20 µs and the smallest frequency difference 49 kHz of signals with a −3 dB bandwidth of 40 kHz. Among the tested methods, the Reassigned Smoothed Pseudo Wigner-Ville distribution technique was found to be the most capable of localizing closely spaced signal components.

Intra-click time-frequency patterns across the echolocation beam of a beluga whale

The echolocation beam of toothed whales has been studied ever since it was first discovered in 1960. Recent studies have focused on the frequency distributions across the cross sections of the beams. Other studies have focussed on describing the entire acoustic field around the animal. However, no one has yet described the timing of each frequency component in the main lobe beam in relation to the other frequency components. Even though the echolocation clicks of broadband click species like the beluga whales (Delphinapterus leucas) are short in time (around 70 μs), previous results have shown indications on a frequency dependence with time, within each click. Little is known about the details of how the signal is generated and transmitted into the water. Investigations of when in time the frequency components occur within each click would give us further knowledge to how the signals are generated. This study takes a closer look at these intra-click time-frequency patterns across the echolocation beam of ...

Inference for time-varying signals using locally stationary processes

Abstract Locally Stationary Processes (LSPs) in Silverman’s sense, defined by the modulation in time of a stationary covariance function, are valuable in stochastic modelling of time-varying signals. However, for practical applications, methods to conduct reliable parameter inference from measured data are required. In this paper, we address the lack of suitable methods for estimating the parameters of the LSP model, by proposing a novel inference method. The proposed method is based on the separation of the two factors defining the LSP covariance function, in order to take advantage of their individual structure and divide the inference problem into two simpler sub-problems. The method’s performance is tested in a simulation study and compared with traditional sample covariance based estimation. An illustrative example of parameter estimation from EEG data, measured during a memory encoding task, is provided.