Architecture adjusts for wireless voice
01 October 2006
Intercepting voice communications from wireless transmissions is a challenging task. System architecture can collect voice signals in the VHF/UHF range with minimum operator attendance.
The RF spectrum is crowded with intermittent and short-duration transmissions. Spectrum monitoring systems look at an expanse of the frequency range and are sometimes called signal survey or emitter identification systems. They may also have a narrowband mode to further process signal data. Voice signals are a specific type of signal transmission that can be selected in some signal monitoring solutions.
Identifying voice transmissions
Some of the challenges in voice interception are an increasing number of wireless transmission devices, plus the number and complexity of transmission protocols. The newest protocols make it very difficult to distinguish voice from noise.
The wireless transmissions containing voice can be over a broad frequency range and at unknown frequencies. Some signal monitoring systems may sweep the broad spectrum too slowly to find many of the short-duration transmissions. One solution is to use multiple narrowband receivers, that could significantly increase hardware and operator costs.
Voice signals may not be transmitting consistently and some voice monitoring systems can be tedious to run and operator intensive to find signals of interest. Determining signals of interest from the thousands of transmissions often wastes time and resources.
System components
The complete architecture for collecting voice transmissions contains numerous components or subsystems. Front-end hardware maximises the probability of intercepting the voice signals. The back-end processing and recording of the voice signals provides the critical information. However, information on too many extraneous signals wastes resources. The ease of integrating front-end and back-end processing with other sub-systems, e.g. direction finding, should also be considered. When making decisions on the solution architecture equipment costs should be considered with the cost of operation, deployment, maintenance and potential for upgrades.
Many systems are based upon a narrowband receiver front-end or multiple narrowband receivers. In simple systems, an operator manually tunes the receiver to a predetermined frequency list, or scans a designated range of frequencies. Recording signals of interest is also initiated by the operator, although help from a reviewer can determine if a particular signal should be collected.
Some narrowband receivers can automatically scan a frequency range or specific frequencies. The scan automatically stops when the receiver detects energy over a pre-determined signal level. It still relies on a full-time operator/reviewer to determine if the signal is voice, and then initiate recording.
Each narrowband receiver covers a fraction of the frequency range of interest, so signals outside of the narrowband will be missed. To increase the probability signals are captured; multiple narrowband receivers cover more of the range of interest (see figure 1). However, complete coverage of a wide frequency range using multiple narrowband receivers becomes costprohibitive.
Data collection
In automated systems with multiple narrowband receivers, large amounts of data are collected in a host-processing system. These are designed with high speed interfaces from the receivers and usually have massive amounts of data storage.
The narrowband receiver architecture is a low capital cost solution for simple voice monitoring in a narrow frequency band. One drawback is the need for operator/reviewer attention to capture the signals of interest.
To automate this architecture requires significantly more equipment and resources for the custom programming. Even then, the system created may not maximise the probability of capturing signals of interest because generally only one signal in a narrowband is recorded.
A wideband search with narrowband collection system improves voice intercept architecture. The wideband search detects candidate voice transmissions in the wideband spectrum, and then ‘tips’ the narrowband voice processing. Confirmed voice signals from the system are recorded to files for offline processing by one or more reviewers. This system architecture allows one or two narrowband channels to be listened to in real-time.
A system with very fast wideband search would be able to measure signals that are present for less than a second. To find small signals close to large signals or to find small signals near to the noise floor, high resolution is needed. The wideband search system should be able to monitor a 1GHz spectrum with a 2kHz resolution bandwidth in 250msecs. Users can discover unknown short-duration signals with a wideband, highspeed, high-resolution front-end.
Unwanted signals
With many transmissions in the crowded RF spectrum, an automated technique discovers signals of interest and records them. An advanced voice intercept system would first use thresholds for defining signal levels of interest. If the spectral energy breaks the threshold criteria, that would result in a signal detection. For each signal above the threshold, information is captured and recorded in a real-time database to be used with automated detection tools.
Alarms and alarm tasks would automate data gathering and define voice signals of interest. Certain signals can be ignored once determined irrelevant. Voice monitoring systems generally require a large number of operators, so creating an unattended system using alarms and tasks saves money.
In this system architecture, the voice processing is done in a narrowband mode. The search engine is positioned to a 36MHz span and the multiple narrowband channels are activated for recording voice simultaneously. The narrowband signals are extracted from the wideband data stream using a bank of digital down converters. Typically 32 to 92 narrowband channels provide adequate capacity for monitoring voice signals in the 36MHz span. Narrowband channels can have programmable bandwidths to optimise the captured voice signals.
These narrowband channels substitute for purchasing external handoff receivers because you can lock one of these channels to a frequency of interest. This integration of handoff receivers with the voice processing system makes it much easier to collect the signal data in the file processing system. Most external handoff receivers provide only analogue audio output.
Significant DSP (digital signal processing) horsepower is required for voice detection and analysis. For voice intercepts, DSP tasks would include identifying FM-modulated signals, demodulating the signal, and detecting voice with special processing algorithms. Signal data is formatted and buffered for output to the host PC for recording to disk. Recordings should continue until the end of transmission, userspecified limits are met, or an out-of-range LOB from the DF sub-system.
Listening to recordings
Efficient post-processing can minimise the number of reviewers. Both a real-time audio output and an audio player would optimise the use of operators and reviewers.
Operators can listen to one or two narrowband channels in real time. This audio output mode employs the same narrowband processing infrastructure as used by the voice processing system, thus reducing the need for separate handoff receivers. The audio data is streamed to the host workstation, where it is heard using the computer’s sound system.
Review operation must be asynchronous to the voice intercept system’s recording of files, so information is not lost. To maximise throughput requires full-featured audio playback software, plus superior processing methodology.
Software tools should have key playback features along with advanced demodulation features. For example, a squelch function ‘skips over’ portions of the recording that do not contain voice. Now, playback time is dramatically compressed for many two-sided voice transmissions.
Efficient file management is another key feature. The review process should be a hierarchical structure with recordings sorted into folders or deleted. Multiple reviewers may be needed to monitor a single directory of recordings. Reviewing should be while the voice intercept system asynchronously provides more data files for review. Links between the audio player and the voice intercept system will continue to update an ‘ignore list’ of frequencies.
Integrating sub-systems
Front-end architectures can be integrated with a direction finding system, so a line-ofbearing (LOB) can be identified on voice transmissions. The LOB information on signals may determine if it is of interest or not. Software should allow sockets interfaces and have application programming interfaces (API) so system developers can customise the user interface or add other hardware and software.
The advanced voice collection architecture has been considered in the design of the E3238S Signal Monitoring System. Hardware is configured for monitoring signals in the VHF/UHF range. Its wideband search capability with high speed and high resolution identifies and determines the unknown, intermittent candidate signals. The optional narrowband voice processing using the proprietary 35688E-VA2 software algorithm collects information on only the candidate signals that contain voice.
CHRIS DESALVO is product marketing engineer, instrument systems, Agilent Technologies.
Contact Details and Archive...
Related Articles...
Most Viewed Articles...