When using frequency-modulated modulation methods, such as those more generally found in VHF and UHF speech communications, it is electrically easy to discriminate between the wanted radio signal and the unwanted background noise. Such modes generate a constant transmit carrier signal, indicating that a station is present. When the carrier signal of a station can no longer be detected by a radio receiver, the circuitry is capable of detecting this and mutes the loudspeaker audio so that the listener is not disturbed by the high levels of background noise. This facility is known as SQUELCH.
In order for a gateway to switch between receive and transmit (necessary to allow two-way communication over a single channel), the status of the radio receiver squelch must be known. Most modern receivers provide a special output, the voltage of which changes depending on whether the receiver squelch is OPEN (a station is being received and audio is heard) or CLOSED (the receiver is muted because no station is being received). In simple terms, this voltage output can be applied to a computer port. With appropriate software, the computer can detect whether the radio receiver is active or dormant by monitoring changes in the squelch voltage.
However, the modulation type to be used in this research is 'single sideband', an amplitude-modulated mode of transmission. There is no constant carrier signal with this mode. The signal comes and goes in sympathy with the patterns of the human voice. It is, of course, still possible to apply the squelch technique described above to these signals. When the sending station speaks, a signal is detected by the receiver and the squelch opens.
However, shortwave frequencies are highly prone to all manner of non-speech signals too. These can include:
* Man-made interference, such as switching clicks and noise generated by TV sets, computers, modems, power tools, thermostats and other electrical equipment;
* Natural interference, such as lightning static, discharges and other solar noise;
* Non-speech transmission modes, such as morse code, teleprinter, radio FAX and other data signals;
* Stations that are aligning (tuning) their transmitters, which often manifests as a constant tone or tones;
* Bleed from human speech signals that are close to, but not actually on, the channel being monitored.
The problem arises that all of the above, often in combination, are radio signals that will open a standard squelch. The squelch thus far described cannot discriminate between the wanted speech signal and the unwanted interference. What is now required is an INTELLIGENT SQUELCH, one that can recognise human speech and only opens the receiver when it has been detected.
The recognition of the characterstic sound signal of human speech is complicated. Detection has to take into account the spectral make-up of the signal, from bass to high treble sections. The syllabic or rythmic nature of the spoken word is a factor, as is the pattern of percussive and sibilant elements of the vocal range.
In isolating what is human speech and what is not, variations in the technical quality of received audio and the differences in patterns of speech between different languages, dialects and nationality must not be a factor in effective detection.
Military developers and a few civilian circuit designers have produced highly complex squelch circuits for use in such trying radio environments. Circuits can be downloaded from the Internet and it is the intention to explore the effectiveness of these systems as part of this research.
It is, however, noteable that the solutions found so far all appear to require the construction of electrical circuits, with fixed components on a circuit board. It seems so much more flexible to create a speech-recognition system in software terms; that is, a computer program that can carry out this task using DSP (digital signal processing). Such a facility might allow 'fine tuning' of its operation by adjusting key parameters and observing the change in performance that results. Bear in mind that the task is not to understand the words being said, simply that the signal is confirmable as human conversation.
The Gateway at G4HPE-L has now been fitted with the excellent VOS-4 intelligent squelch circuit available from Naval Communications in the USA. This is still under assessment but appears to work very well. It seems to recognise speech and ignores noise, tuning transmitters and data transmissions. A more detailed report on this system will appear here soon.
If you are someone who has an interest in developing suitable software, or you have come across such an item, the research programme would be very pleased to hear from you!