Echolocation is the ability to use sound-echoes to infer spatial information about the environment. Some blind people have developed extraordinary proficiency in echolocation using mouth-clicks. The first step of human biosonar is the transmission (mouth click) and subsequent reception of the resultant sound through the ear. Existing head-related transfer function (HRTF) data bases provide descriptions of reception of the resultant sound. For the current report, we collected a large database of click emissions with three blind people expertly trained in echolocation, which allowed us to perform unprecedented analyses. Specifically, the current report provides the first ever description of the spatial distribution (i.e. beam pattern) of human expert echolocation transmissions, as well as spectro-temporal descriptions at a level of detail not available before. Our data show that transmission levels are fairly constant within a 60° cone emanating from the mouth, but levels drop gradually at further angles, more than for speech. In terms of spectro-temporal features, our data show that emissions are consistently very brief (~3ms duration) with peak frequencies 2-4kHz, but with energy also at 10kHz. This differs from previous reports of durations 3-15ms and peak frequencies 2-8kHz, which were based on less detailed measurements. Based on our measurements we propose to model transmissions as sum of monotones modulated by a decaying exponential, with angular attenuation by a modified cardioid. We provide model parameters for each echolocator. These results are a step towards developing computational models of human biosonar. For example, in bats, spatial and spectro-temporal features of emissions have been used to derive and test model based hypotheses about behaviour. The data we present here suggest similar research opportunities within the context of human echolocation. Relatedly, the data are a basis to develop synthetic models of human echolocation that could be virtual (i.e. simulated) or real (i.e. loudspeaker, microphones), and which will help understanding the link between physical principles and human behaviour. Echolocation is the ability to use sound-echoes to infer spatial information about the environment. It is well known from certain species of bats or marine mammals. Remarkably, some blind people have developed extraordinary proficiency in echolocation using mouth-clicks. Human echolocation work has built on scant theoretical foundations to date. The current report characterizes the transmission (i.e. mouth click) that people use for echolocation, and in this way provides data that can be used to advance the field in a theory guided way. We collected a large database of mouth clicks with three blind people expertly trained in echolocation. This allowed us to perform unprecedented analyses. Specifically, the current report provides the first ever description of the beam pattern of human expert echolocation transmissions, as well as spectro-temporal descriptions at a level of detail not available before. Based on our measurements we also propose a mathematical model to synthesize transmissions. Thus, the data are a basis to develop synthetic models of human echolocation, which are essential for understanding characteristics of click echoes and human echolocation behaviour in tasks such as localising or recognising an object, navigating around it etc.

The automobile industry has a strategic intent to develop vehicles capable of higher and higher levels of autonomy. Radar is a key sensing component, being the only viable option for all weather, 24-hour-a-day use. Autonomy will require the radar to provide much greater levels of detailed information than is available from today's cruise control and collision avoidance systems. In particular, much wider bandwidths are needed for high range resolution. Furthermore, high Doppler resolution, requiring long integration times, is needed for moving object classification. These two opposing requirements create significant design and signal processing challenges. This paper examines the way in which choices in the key design parameters effect performance, showing that as both range and Doppler resolutions are enhanced, the potential for received energy to migrate across resolution cells increases and must be corrected to enable measurement of both focused high-resolution imagery and micro-Doppler signatures, as well as restoring system sensitivity. An advanced range and Doppler cell migration correction (RDCMC) algorithm is introduced and performance evaluated using both theory and experimentation.

In this article, a method for creating highly realistic synthetic drone micro-Doppler spectrograms is presented and its effectiveness of training a bird-drone classifier for real scenario classification is shown via comparisons to a real benchmark. The effect of drone motor speed sampling used when simulating drone micro-Doppler is shown to have a significant impact on the accuracy of synthetic results and variations of this approach are explored. Four synthetic datasets were created differing in motor speed sampling and each were compared in their ability to train a convolutional neural network to classify real data. The highest fidelity synthetic dataset achieved a classification accuracy of 86.6% compared to the real benchmark accuracy of 89.7%. The adverse effect on classifier robustness when reducing the simulation fidelity by altering the motor speed sampling is shown.

This article develops and examines techniques to enable coherent change detection (CCD) for short-range, high-resolution drone-borne synthetic aperture radar (SAR) systems operating at high frequencies. The potential of using high frequencies at short ranges for fine-resolution imagery and sensitivity to temporal change detection is highlighted, as are the challenges in terms of sensitivity to SAR motion errors. SAR system characteristics for CCD are derived, and the impact of motion errors, which leads to spatial decorrelation and co-registration errors on CCD maps, is discussed. Subsequently, a CCD algorithm able to generate change maps with a >0.75 average coherence value is presented. The validity of the approach is tested through various experimental scenarios. As a result, car tyre marks and human footprints are possible to discern with a drone-borne SAR demonstrator operating at 24 GHz.

Passive bistatic radar (PBR), exploiting broadcast, communications or radionavigation signals, has received a great deal of attention in recent years, and has a number of attractions compared with conventional radar systems. However, the waveforms are not fundamentally designed for radar operation, so their performance will in general be suboptimal. It is therefore important to understand the effect of the waveform on the performance of the PBR, so as to be able to choose the most appropriate illuminator and to use the waveform in the optimal way, and it is in this sense that PBR forms a subject of waveform diversity. This chapter reviews the properties of a range of different PBR waveforms and the processing methods used with them.

Deep learning with convolutional neural networks has been widely utilised in radar research concerning automatic target recognition. Maximising numerical metrics to gauge the performance of such algorithms does not necessarily correspond to model robustness against untested targets, nor does it lead to improved model interpretability. Approaches designed to explain the mechanisms behind the operation of a classifier on radar data are proliferating, but bring with them a significant computational and analysis overhead. This work uses an elementary unsupervised convolutional autoencoder (CAE) to learn a compressed representation of a challenging dataset of urban bird and drone targets and subsequently if apparent quality of the representation via preservation of class labels leads to better classification performance after a separate supervised training stage. It is shown that a CAE that reduces the features output after each layer of the encoder gives rise to the best drone vs bird classifier. A clear connection between unsupervised evaluation via label preservation in the latent space and subsequent classification accuracy after supervised fine-tuning is shown, supporting further efforts to optimise radar data latent representations to enable optimal performance and model interpretability.