This PDF file contains the front matter associated with SPIE Proceedings Volume 7305, including the Title Page, Copyright information, Table of Contents, and the Conference Committee listing.

Most of Japan's local governments utilize municipal disaster-management radio communications systems to communicate information on disasters or terrorism to residents. The national government is progressing in efforts toward digitalization by local governments of these systems, but only a small number (approx. 10%) have introduced such equipment due to its requiring large amounts of investment. On the other hand, many local governments are moving forward in installation of optical fiber networks for the purpose of eliminating the "digital divide." We herein propose a communication system as an alternative or supplement to municipal disaster-management radio communications systems, which utilizes municipal optical fiber networks, the internet and similar networks and terminals. The system utilizes the multiple existing networks and is capable of instantly distributing to all residents, and controlling, risk management information. We describe the system overview and the field trials conducted with a local government using this system.

At a border security conference in August 2008, Michael Sullivan, acting director of Bureau of Alcohol, Tobacco, Firearms and Explosives stated that, "Nearly all illegal firearms (90% to 95%) seized in Mexico come from the United States"[1]. When firearms are recovered at a crime scene, the firearms can be traced providing specific details on illegal firearm dealers or straw purchasers within the United States. Criminals or narco terrorist groups target US dealers to source firearms for drug cartels in Mexico and South America. Joint law enforcement programs between the US and Mexico law enforcement have been effective, however, in most cases the firearms that are seized are only a small fraction of the firearms trafficked across the United States border. A technology called Microstamping, when applied to newly manufactured firearms will provide further opportunities for tracing illegal firearms for law enforcement in the United States and across the globe. Microstamping is a patented technology and trace solution where intentional tooling marks are formed or micromachined onto firearms interior surfaces that come into contact or impact the surfaces of cartridge casings. The intentional tooling marks can take the form of alphanumeric codes or encoded geometric codes, such as a barcode. As the firearm is discharged the intentional tooling marks transfer a code to the cartridge casing before it is ejected out of the firearm. When recovered at the scene of an incident, the Microstamped cartridge can indentify a specific firearm, without the need to recover that firearm. Microstamping provides critical intelligence for use in border security operations and cross border violent drug related crime investigations. This paper will explain the key attributes of microstamping technology; including its potential benefits in border security operations and how data gathered from the technique can be used in geospatial information systems to identify illicit firearm sources, trafficking routes, as well as spatial and temporal mapping of narco-terrorist movements on either side of the border.

Warfare relies on effective, accurate and timely intelligence an especially critical task when conducting a counterinsurgency operation [1]. Simply stated counterinsurgency is an intelligence war. Both insurgents and counterinsurgents need effective intelligence capabilities to be successful. Insurgents and counterinsurgents therefore attempt to create and maintain intelligence networks and fight continuously to neutralize each other's intelligence capabilities [1][2]. In such an environment it is obviously an advantage to target or proactively create opportunities to track and map an insurgent movement. Quickly identifying insurgency intelligence assets (Infiltrators) within a host government's infrastructure is the goal. Infiltrators can occupy various areas of government such as security personnel, national police force, government offices or military units. Intentional Firearm Microstamping offers such opportunities when implemented into firearms. Outfitted within firearms purchased and distributed to the host nation's security forces (civilian and military), Intentional Firearm Microstamping (IFM) marks bullet cartridge casings with codes as they are fired from the firearm. IFM is incorporated onto optimum surfaces with the firearm mechanism. The intentional microstamp tooling marks can take the form of alphanumeric codes or encoded geometric codes that identify the firearm. As the firearm is discharged the intentional tooling marks transfer a code to the cartridge casing which is ejected out of the firearm. When recovered at the scene of a firefight or engagement, the technology will provide forensic intelligence allowing the mapping and tracking of small arms traffic patterns within the host nation or identify insurgency force strength and pinpoint firearm sources, such as corrupt/rogue military units or police force. Intentional Firearm Microstamping is a passive mechanical trace technology that can be outfitted or retrofitted to semiautomatic handguns and military rifles to assist in developing real time intelligence providing a greater level of situational awareness. Proactively Microstamping firearms that are introduced and distributed to the host nation's security forces, it will become easier to track the firearms if they go missing or end up on the black market in the hands of an insurgency. This paper will explain the technology and key attributes of microstamping technology, test data showing its ability to identifying a specific firearm, examples of implementation strategies and to what extent data could be utilized in war zone security and counterinsurgency intelligence operations.

This paper pertains to the reduction of measurement errors due to drift in rate gyros used for tracking the position of moving vehicles. In these applications, gyros and odometry are often used to augment GPS when GPS reception is unavailable. Drift in gyros causes the unbounded growth of errors in the estimation of heading, rendering low-cost gyros almost entirely useless in applications that require good accuracy for more than just a few seconds or minutes. Our proposed method, called "Heuristic Drift Reduction" (HDR), applies a unique closed-loop control system approach to estimate drift in real-time and remove the estimated drift instantaneously from the gyro reading. The paper presents results of experiments, in which a gyro-equipped car was driven hundreds of miles on highways, rural roads, and city streets. HDR reduced the average heading error over all of these drives by one order of magnitude.

CBRN Crime Scene Modeler (C2SM) is a prototype mobile CBRN mapping system for First Responders in events where Chemical, Biological, Radiological and Nuclear agents where used. The prototype operates on board a small robotic platform, increases situational awareness of the robot operator by providing geo-located images and data, and current robot location. The sensor suite includes stereo and high resolution cameras, a long wave infra red (thermal) camera and gamma and chemical detectors. The system collects and sends geo-located data to a remote command post in near real-time and automatically creates 3D photorealistic model augmented with CBRN measurements. Two prototypes have been successfully tested in field trials and a fully ruggedised commercial version is expected in 2010.

This paper presents a distributed multi-modality sensor network concept for vehicle classification within perimeter of a surveillance system. This perimeter surveillance concept represents a "Virtual RF Fence" consisting of remotely located electro-optic surveillance cameras and a standoff range radar system. The perimeter surveillance system vigilantly monitors the field and each time a vehicle crosses the virtual RF fence it informs the surveillance cameras to actively monitor the activity of vehicles as it passes through the field. This paper describes the methodologies applied for processing the EO imagery data including target vehicle segmentation from background, vehicle shadow elimination, vehicle feature vector generation, and a neural network approach for vehicle classification. A metric is also proposed for evaluation of performance of the vehicle classification technique.

In a harbor environment threats like explosives-packed rubber boats, mine-carrying swimmers and divers must be detected in an early stage. This paper describes the integration and use of a heterogeneous multiple camera system with panoramic observation capabilities for detecting these small vessels in the Den Helder New Harbor in the Netherlands. Results of a series of experiments with different targets are presented. An outlook to a future sensor package containing panoramic vision is discussed.

Helicopters present a serious threat to high security facilities such as prisons, nuclear sites, armories, and VIP compounds. They have the ability to instantly bypass conventional security measures focused on ground threats such as fences, check-points, and intrusion sensors. Leveraging the strong acoustic signature inherent in all helicopters, this system would automatically detect, classify, and accurately track helicopters using multi-node acoustic sensor fusion. An alert would be generated once the threat entered a predefined 3-dimension security zone in time for security personnel to repel the assault. In addition the system can precisely identify the landing point on the facility grounds.

Maintaining tactical superiority in a complex battlespace, with asymmetric threats, dictates the need for a real-time, high throughput imaging system that provides the user with an intuitive and effective means of achieving imaging situational awareness quickly and continuously. As systems and sensors grow in both number and intricacy, situational awareness necessitates presenting large amounts of data in an easily understandable manner to the operator to avoid fatigue and overload. This paper discusses how we achieved real-time, 360° imaging awareness and presents a demonstrated prototype. The imaging system architecture is described with emphasis on the dedicated high-speed image processor which provides the real-time panoramic stitching, digital zoom and image stabilization. The image processor is designed around a high-speed, state-of-the-art Commercial-off-the-shelf (COTS) FPGA architecture that executes algorithms on these high speed digital video streams with less than 100ms of latency. Field test results are presented from recent at-sea and ground vehicle data collects. Planned future expansions to the system are outlined such as the capacity for higher resolution, infrared capability, and the planned platform integration efforts. Other important lessons learned as a result of this development are also presented throughout the paper.

Protecting national borders, military and industrial complexes, national Infrastructure and high-value targets is critical to national security. Traditional solutions use a combination of ground surveillance radar, motion detection systems and video surveillance systems. Our development objective was to provide wide area 360-degree surveillance and ground-moving target detection using a passive optical system. In order to meet this objective, the development of an optical system capable of wide-area surveillance with intelligent cueing, high-resolution tracking and target identification is required. The predominant approach to optical surveillance has traditionally been gimbaled narrow field-of-view systems. These systems miss the majority of events occurring around them because of their inability to focus on anything other than a single event or object at any one time. Details of the system requirements definition, design trade studies and selected design configurations are discussed. The experimental results obtained during the current development phase have provided consistently high quality images and enhanced situational awareness. A summary of field validation methods and results is provided.

The aim of maritime surveillance systems is to detect threats early enough to take appropriate actions. We present the results of a study on maritime domain awareness performed during the fall of 2008. We analyze an identified capability gap of worldwide surveillance in the maritime domain, and report from a user workshop addressing the identified gap. We describe a SMARTracIn concept system that integrates information from surveillance systems with background knowledge on normal conditions to help users detect and visualize anomalies in vessel traffic. Land-based systems that cover the coastal waters as well as airborne, space-borne and ships covering open sea are considered. Sensor data are combined with intelligence information from ship reporting systems and databases. We describe how information fusion, anomaly detection and semantic technology can be used to help users achieve more detailed maritime domain awareness. Human operators are a vital part of this system and should be active components in the fusion process. We focus on the problem of detecting anomalous behavior in ocean-going traffic, and a room and door segmentation concept to achieve this. This requires the ability to identify vessels that enter into areas covered by sensors as well as the use of information management systems that allow us to quickly find all relevant information.

In this paper we present a new particle filter based multi-target tracking method incorporating Gaussian Process Dynamical Model (GPDM) to improve robustness in multi-target tracking on complex motion patterns. With the Particle Filter Gaussian Process Dynamical Model (PFGPDM), a high-dimensional training target trajectory dataset of the observation space is projected to a low-dimensional latent space through Probabilistic Principal Component Analysis (PPCA), which will then be used to classify test object trajectories, predict the next motion state, and provide Gaussian process dynamical samples for the particle filter. In addition, histogram- Bhartacharyya and GMM Kullback-Leibler are employed respectively, and compared in the particle filter as complimentary features to coordinate data used in GPDM. Experimental tests are conducted on the PETS2007 benchmark dataset. The test results demonstrate that the approach can track more than four targets with reasonable run-time overhead and good performance.

The SBInet program to secure our border with physical fences and electronic surveillance uses Unattended Ground Sensors (UGS) to detect illegal crossing activity. The presentation will discuss the role of UGS in SBInet for cueing tower based surveillance systems and as an independent surveillance system in areas where tower surveillance is impractical. The presentation will provide a status of UGS integration into the SBInet Common Operating Picture (COP). McQ as the supplier of UGS for SBInet has supported the system integration with radars and long range imagers. McQ has worked with DHS and the Border Patrol in refining the UGS surveillance application and in training for their use. The presentation will address new UGS technology for border monitoring.

Binary sensor systems are various types of analog sensors (optical, MEMS, X-ray, gamma-ray, acoustic, electronic, etc.), based on the binary decision process. Typical examples of such "binary sensors" are X-ray luggage inspection systems, product quality control systems, automatic target recognition systems, numerous medical diagnostic systems, and many others. In all these systems, the binary decision process provides only two mutually exclusive responses: "signal" and "noise." There are also two types of key parameters that characterize either system (such as false positive and false negative), or a priori external-to-system conditions (such as absolute probabilities). In this paper, by using a strong medical analogy, we analyze a third type of key parameter that combines both system-like and a priori information, in the form of so called Bayesian Figures of Merit, and we show that the latter parameter, in the best way, characterizes a binary sensor system.

System-on-chip (SoC) single-die electronic integrated circuit (IC) integration has recently been attracting a great deal of attention, due to its high modularity, universality, and relatively low fabrication cost. The SoC also has low power consumption and it is naturally suited to being a base for integration of embedded sensors. Such sensors can run unattended, and can be either commercial off-the-shelf (COTS) electronic, COTS microelectromechanical systems (MEMS), or optical-COTS or produced in house (i.e., at Physical Optics Corporation, POC). In the version with the simplest electronic packaging, they can be integrated with low-power wireless RF that can communicate with a central processing unit (CPU) integrated in-house and installed on the specific platform of interest. Such a platform can be a human body (for e-clothing), unmanned aerial vehicle (UAV), unmanned ground vehicle (UGV), or many others. In this paper we discuss SoC-centric embedded unattended sensors in Homeland Security and military applications, including specific application scenarios (or CONOPS). In one specific example, we analyze an embedded polarization optical sensor produced in house, including generalized Lambertian light-emitting diode (LED) sources and secondary nonimaging optics (NIO).

Unattended Ground Sensors (UGS) have recently gained momentum in surveillance and protection applications. Many of these Unattended Ground Sensors are deployed in current operations today across the Department of Defense (DoD) and Department of Homeland Security (DHS). In addition to UGS needs, there is a growing desire to leverage existing UGS for incorporation into higher level systems for a broadening role in defense and homeland security applications. The architecture to achieve this goal and examples of non-traditional scenarios that leverage higher level systems are discussed in this paper.

Battery power resource management becomes a critical issue in the case of self-powered remote wireless RF electronics, where the basic parameter is time of system operation before battery recharging or battery replacement. In such cases, very often related to physical protection against antitampering (AT), proper theoretical modeling of a battery driven power supply in the context of a given digital electronic system is of utmost importance. Such modeling should include various types of batteries (primary and secondary), various self-discharge processes in different temperatures, and even energy harvesting, the latter to supply power for long-term content, low-power electronic subsystems. In this paper we analyze simple modeling of resource power management, including variations of all of these parameters and energy harvesting.

Combat resiliency is the ability of a commander to prosecute, control, and consolidate his/her's sphere of influence in adverse and changing conditions. To support this, an infrastructure must exist that allows the commander to view the world in varying degrees of granularity with sufficient levels of detail to permit confidence estimates to be levied against decisions and course of actions. An infrastructure such as this will include the ability to effectively communicate context and relevance within and across the battle space. To achieve this will require careful thought, planning, and understanding of a network and its capacity limitations in post-event command and control. Relevance and impact on any existing infrastructure must be fully understood prior to deployment to exploit the system's full capacity and capabilities. In this view, the combat communication network is considered an integral part of or National communication network and infrastructure. This paper will describe an analytical tool set developed at ORNL and RNI incorporating complexity theory, advanced communications modeling, simulation, and visualization technologies that could be used as a pre-planning tool or post event reasoning application to support response and containment.

Rapid improvements in communications infrastructure and sophistication of commercial hand-held devices provide a major new source of information for assessing extreme situations such as environmental crises. In particular, ad hoc collections of humans can act as "soft sensors" to augment data collected by traditional sensors in a net-centric environment (in effect, "crowd-sourcing" observational data). A need exists to understand how to task such soft sensors, characterize their performance and fuse the data with traditional data sources. In order to quantitatively study such situations, as well as study distributed decision-making, we have developed an Extreme Events Laboratory (EEL) at The Pennsylvania State University. This facility provides a network-centric, collaborative situation assessment and decision-making capability by supporting experiments involving human observers, distributed decision making and cognition, and crisis management. The EEL spans the information chain from energy detection via sensors, human observations, signal and image processing, pattern recognition, statistical estimation, multi-sensor data fusion, visualization and analytics, and modeling and simulation. The EEL command center combines COTS and custom collaboration tools in innovative ways, providing capabilities such as geo-spatial visualization and dynamic mash-ups of multiple data sources. This paper describes the EEL and several on-going human-in-the-loop experiments aimed at understanding the new collective observation and analysis landscape.

The U.S. Department of Homeland Security's Standoff Technology Integration and Demonstration Program is designed to accelerate the development and integration of technologies, concepts of operations, and training to defeat explosives attacks at large public events and mass transit facilities. The program will address threats posed by suicide bombers, vehicle-borne improvised explosive devices, and leave-behind bombs. The program is focused on developing and testing explosives countermeasure architectures using commercial off-the-shelf and near-commercial standoff and remotely operated detection technologies in prototypic operational environments. An important part of the program is the integration of multiple technologies and systems to protect against a wider range of threats, improve countermeasure performance, increase the distance from the venue at which screening is conducted, and reduce staffing requirements. The program will routinely conduct tests in public venues involving successively more advanced technology, higher levels of system integration, and more complex scenarios. This paper describes the initial field test of an integrated countermeasure system that included infrared, millimeter-wave, and video analytics technologies for detecting person-borne improvised explosive devices at a public arena. The test results are being used to develop a concept for the next generation of integrated countermeasures, to refine technical and operational requirements for architectures and technologies, and engage industry and academia in solution development.

We apply a unique hierarchical audio classification technique to weapon identification using gunshot analysis. The Audio Classification classifies each audio segment as one of ten weapon classes (e.g., 9mm, 22, shotgun etc.) using lowcomplexity Gaussian Mixture Models (GMM). The first level of hierarchy consists of classification into broad weapons categories such as Rifle, Hand-Gun etc. and the second consists of classification into specific weapons such as 9mm, 357 etc. Our experiments have yielded over 90% classification accuracy at the coarse (rifle-handgun) level of the classification hierarchy and over 85% accuracy at the finer level (weapon category such as 9mm).

Acoustic detection of gunshots has many security and military applications. Most gunfire produces both an acoustic muzzle-blast signal as well as a high-frequency shockwave. However some guns do not propel bullets with the speed required to cause shockwaves, and the use of a silencer can significantly reduce the energy of muzzle blasts; thus, although most existing commercial and military gunshot detection systems are based on shockwave detection, reliable detection across a wide range of applications requires the development of techniques which incorporate both muzzle-blast and shockwave phenomenologies. The detection of muzzle blasts is often difficult due to the presence of non-stationary background signals. Previous approaches to muzzle blast detection have applied pattern recognition techniques without specifically considering the non-stationary nature of the background signals and thus these techniques may perform poorly under realistic operating conditions. This research focuses on time domain modeling of the non-stationary background using Bayesian auto-regressive models. Bayesian parameter estimation can provide a principled approach to non-stationary modeling while also eliminating the stability concerns associated with standard adaptive procedures. Our proposed approach is tested on a synthetic dataset derived from recordings of actual background signals and a database of isolated gunfire. Detection results are compared to a standard adaptive approach, the least-mean squares (LMS) algorithm, across several signal to background ratios in both indoor and outdoor conditions.

A new approach to gunfire location coupling antenna design to field models and signal processing procedures enables direction finding and ranging of projectile sources in spectrally competitive environments, the ranging permitted in certain circumstances. The approach is based upon the notion that data collection should enable mathematical models for incident acoustic fields in antenna neighborhoods, permitting utilization of systems having high resolving power. Theory, procedures, and design are outlined and gunfire location field test results incorporating multiple shooters, echoes, and reverberation are presented. *Technology protected by US Patents 7,423,934; 7,394,724;,7,372,774; 7,123,548; and patents pending.

Increasing battlefield awareness can improve both the effectiveness and timeliness of response in hostile military situations. A system that processes acoustic data is proposed to handle a variety of possible applications. The front-end of the existing biomimetic acoustic direction finding system, a mammalian peripheral auditory system model, provides the back-end system with what amounts to spike trains. The back-end system consists of individual algorithms tailored to extract specific information. The back-end algorithms are transportable to FPGA platforms and other general-purpose computers. The algorithms can be modified for use with both fixed and mobile, existing sensor platforms. Currently, gunfire classification and localization algorithms based on both neural networks and pitch are being developed and tested. The neural network model is trained under supervised learning to differentiate and trace various gunfire acoustic signatures and reduce the effect of different frequency responses of microphones on different hardware platforms. The model is being tested against impact and launch acoustic signals of various mortars, supersonic and muzzle-blast of rifle shots, and other weapons. It outperforms the cross-correlation algorithm with regard to computational efficiency, memory requirements, and noise robustness. The spike-based pitch model uses the times between successive spike events to calculate the periodicity of the signal. Differences in the periodicity signatures and comparisons of the overall spike activity are used to classify mortar size and event type. The localization of the gunfire acoustic signals is further computed based on the classification result and the location of microphones and other parameters of the existing hardware platform implementation.

In this paper, we show that objects of interest, like pipes and cylinders, reminiscent of guns and rifles, can be classified based on their acoustic vibration signatures. That is, if the acoustic returns are measurable, one can indeed classify objects based on the physical principle of resonance. We consider classifiers which are both training independent and those that are training dependent. The statistical classifier belongs to the former category, whereas, the neural network classifier belongs to the latter. Comparisons between the two approaches are shown to render both classifiers as suitable classifiers with small classification errors. We use the probability of correct classification as a measure of performance. We demonstrate experimentally that unique features for classification are the resonant frequencies. The measured data are obtained by exciting mechanical vibrations in pipes of different lengths and of different metals, for example, copper, aluminum, and steel, and the measuring of the acoustic returns, using a simple microphone. Autoregressive modeling is applied to the data to extract the respective object features, namely, the vibration frequencies and damping values. We consider two classification problems, 1) Classifying objects comprised of different metals, and 2) Classifying objects of the same material, but made of different lengths. It is shown that classification performance can be improved by incorporating additional features such as the damping coefficients.

The challenge in detecting behind the wall stationary targets appears when the target is close to the wall and has a frequency response that is fully or partially overlapping with the wall's spectrum. In order to detect such targets, background subtraction is usually applied. The main challenge of using this method is the availability of the empty scene, which is typically unavailable to the user. In this paper, we introduce an adaptive background estimation and subtraction technique, to detect objects behind the wall with the focus on human detection. This technique is based on the architecture of the adaptive side-lobe canceller, where a number of antenna elements are used to form a subarray that captures the background in the main beam, while receiving the incident scatterings from the target in the sidelobes. The output of this subarray is then used as the reference signal to suppress the background components at the output of each sensor, through adaptive Recursive Least Squares (RLS) algorithm. This technique can be used with both co- and cross-polarization returns, in order to further reduce the effect of the background and enhance the detectability of the target.

Beginning in 2010, the U.S. will require that all cargo loaded in passenger aircraft be inspected. This will require more efficient processing of cargo and will have a significant impact on the inspection protocols and business practices of government agencies and the airlines. In this paper, we develop an aviation security cargo inspection queuing simulation model for material flow and accountability that will allow cargo managers to conduct impact studies of current and proposed business practices as they relate to inspection procedures, material flow, and accountability.

Optical synchronous coherent detection is attracting greater attention within the defense and security community because it allows linear recovery both of the amplitude and phase of optical signals. Fiber-based transmission impairments such as chromatic dispersion and polarization mode dispersion can be compensated in the electrical domain. Additionally, synchronous detection offers the potential of improved receiver sensitivity and extended reach versus direct or interferometric detection schemes. 28 Gbaud/112 Gb/s and 42.8 Gbaud transmissions are now being considered in fiber networks worldwide. Due to the lack of broadband high frequency components centered at IF values of 56 GHz and 86 GHz, respectively, the coherent heterodyne approach is not viable for these baud rates. The homodyne approach remains one of the choices available to fully exploit the advantages of synchronous coherent detection at these transmission data rates. In order to implement the homodyne receiver, optical phase locking between the signal and local oscillator laser (LO) is required. Digital approaches for this task rely upon very complex, fast, and high power-consumption chips. A homodyne receiver using an analog approach for phase locking would allow for increased system simplicity at a lower cost. Use of commercial-off-the-shelf (COTS) DFB lasers embedded within the receiver would also increase system feasibility for defense applications. We demonstrate synchronous demodulation of a 42.8 Gbaud signal using an analog optical phase-locked loop. The homodyne system was optimized to use COTS DFB lasers having an aggregate linewidth of ~2 MHz. We also analyze the impact of uncompensated phase noise on receiver performance.

A method of high resolution over-the-horizon radars (OTHRs) using time-reversal is described. The method uses timereversal in the multipath-enriched ionospheric environment in order to achieve the extended virtual aperture. Also, the double-pass conjugate image scanning scheme allows imaging of non-cooperative targets without requiring apriori knowledge of environmental conditions. Initial theoretical and experimental results are provided.

Non-language speech sounds (NLSS) are sounds produced by humans that do not carry linguistic information. Examples of these sounds are coughs, clicks, breaths, and filled pauses such as "uh" and "um" in English. NLSS are prominent in conversational speech, but can be a significant source of errors in speech processing applications. Traditionally, these sounds are ignored by speech endpoint detection algorithms, where speech regions are identified in the audio signal prior to processing. The ability to filter NLSS as a pre-processing step can significantly enhance the performance of many speech processing applications, such as speaker identification, language identification, and automatic speech recognition. In order to be used in all such applications, NLSS detection must be performed without the use of language models that provide knowledge of the phonology and lexical structure of speech. This is especially relevant to situations where the languages used in the audio are not known apriori. We present the results of preliminary experiments using data from American and British English speakers, in which segments of audio are classified as language speech sounds (LSS) or NLSS using a set of acoustic features designed for language-agnostic NLSS detection and a hidden-Markov model (HMM) to model speech generation. The results of these experiments indicate that the features and model used are capable of detection certain types of NLSS, such as breaths and clicks, while detection of other types of NLSS such as filled pauses will require future research.

Reliable communication over hostile environment is desired by both military and civilian parties. In this paper, we propose an enhanced transform domain communication system (ETDCS) with narrow-band interference (NBI) avoidance capability. The basic idea for this system is to synthesize adaptive waveform in the frequency domain by a non-parametric spectral estimator, called Capon's method, at both the transmitter and the receiver to avoid spectrally crowded regions. This approach offers better bit error performance than existing similar systems such as the transform domain communication system (TDCS) that utilizes a parametric autoregressive (AR) spectral estimator, the wavelet domain communication system (WDCS) which uses wavelet domain periodogram, or the enhanced wavelet domain communication system (EWDCS) which employs the evolutionary wavelet spectrum (EWS). Specifically, our proposed ETDCS significantly improves the bit error performance under non-stationary interference such as swepttone interference while achieving consistent bit error performance under stationary interference such as partial band, singletone, and multitone interference. Hence, our proposed ETDCS provides a viable alternative for highly reliable communication in interference rich communication environments.

The theory of compressed sensing (CS) has shown that compressible signals can be accurately reconstructed from a very small set of randomly projected measurements. Sparse representation of the signals plays an important role in the signal reconstruction of compressed sensing. In this paper, we propose to use signal modulation information to obtain a better sparse representation for communication signals in compressed sensing. In our approach, a tree-structured modulation classification system is used to classify five types of signal modulations: Amplitude Modulation (AM), Frequency Modulation (FM), Amplitude Shift Keying (ASK), Frequency Shift Keying (FSK) and Phase Shift Keying (PSK). The tree-structured classification system uses four signal features to classify the five modulation types, and all features are computable in the analog domain. To select a sparse transformation for the input signal, we propose a pre-trained Karhunen-Loeve transform (KLT) based CS, in which a set of KLT transformation matrices is obtained by an offline learning process for all modulation types. In an online real-time process, the modulation information of the input signal is classified and then used to select one of the pre-trained KLT matrices for providing a better sparse representation of the signal for CS-based signal reconstruction. Our experimental results show that our modulation classification technique is effective in identifying the five modulation types of noisy input signals, and our KLT based CS reconstruction has much better performances than Fourier and wavelet packet based CS for the communication signals we tested.