Improving digital radio voice quality
Advances in vocoder technology offer improvements in digital, portable and mobile radio voice quality.
In the ongoing battle between those who break the law and those who enforce it, communications technology has been a potent weapon used on both sides. Back in the days of prohibition, bootleggers in fast cars were more often run down by the radio than by horsepower and daring. Now, crooks use computerized scanners to listen in on police radio communications, and law enforcement agencies are doing their best to stay a step ahead.
Digital technology represents the latest step available to law enforcement officers in their efforts to stay ahead of criminals. In a 1997 study on interoperability sponsored by the National Institute of Justice, 55% of law enforcement agencies planning to upgrade or replace their land mobile radio (LMR) systems within the next 10 years said they preferred a digital system over analog. Digital products are less susceptible to eavesdropping and have better reception over more of the radio system’s coverage area.
As shown in Figure 1 on page 36, digital radios outperform analog radios in fringe and weak-signal areas. These occur at the outer edges of a system’s coverage zone, or in areas where transmission and/or reception is poor because of interference caused by terrain or concrete and steel buildings.
Despite these advantages, however, digital radio has yet to gain widespread acceptance among public safety agencies. One of the primary reasons, according to law enforcement officers, is the less-than-stellar audio quality found in digital portable and mobile radios. Even though the message gets through intelligibly in areas where analog may be garbled, in all first-generation digital radios, it doesn’t sound good to the listener.
A frequent complaint about digital products is that the speaker’s voice sounds tinny and mechanical. The tinny sound is a function of the vocoder, or voi ce coder/decoder, used to translate audio from analog to digital and back to analog. In the process of coding and decoding speech, voice quality is compromised to the point where users often have difficulty recognizing the voice of the person on the other end. That means that a police officer might have to look at the radio’s display just to find out who he’s talking to, which he might not be able to do in a critical situation.
“The lack of voice recognition might be OK in some applications, but in law enforcement, I think it’s unacceptable,” said John Walker, telecommunications engineer for Riverside County, CA. “I’m a believer in digital technology. When the sound quality and fidelity meet my expectations, I’ll be interested.”
Improvements in vocoder technology have potential to eliminate most of the problems with digital audio in land mobile radio. Integration of a high-quality vocoder in a digital radio system and its portable and mobile radios enables police officers and other users to have the most true representation of a speaker’s voice available in a digital system. Specifically, the Improved Multi-Band Excitation (IMBE) vocoder shows promise because of its high ratings in studies. When compared to three other vocoder technologies under consideration for wireless communications, the IMBE vocoder was judged to have the best voice quality under each operating condition, according to an evaluation funded by the Telecommunications Industry Association (TIA).
Vocoders work by using a speech analyzer to convert analog speech waveforms into a digital data stream. The vocoder models a segment of the waveform to estimate, quantify and code the speech before transmission. The data stream also must be compressed to fit within a land mobile radio channel. On the receiving end, the vocoder converts the digital stream into artificial speech sounds through the use of a speech synthesizer.
The effects of channel degradation in digital radio, however, make it difficult to design a vocoder with a near-perfect rendering of the user’s voice. For this reason, a vocoder designed for extremely low bit-error rates (BER), such as those found in landline telephone communications, will not perform well in the harsh environment of wireless two-way radio.
To avoid channel degradation, designers must implement methods of making the vocoder operate effectively at high BER. According to Digital Voice Systems (DVSI), the developer of the IMBE vocoder, and Ericsson Private Radio Systems, which licenses the vocoder for use in many of its radios, several techniques can diminish the effects of high BER found in wireless communications. These methods, used in the vocoder algorithm design process and in the implementation of the vocoder in mobile radios, include balancing coding gain and error persistence, interleaving, error mitigation and forward-error correction.
Achieving the necessary balance between coding gain and error persistence is done by reducing the number of previous frames in speech used to predict the parameters of the current frame. Because the speech waveform varies little from frame to frame, it’s easy to use previous frames to construct the parameters for the current frame. But when channel errors occur, those errors tend to get strung out over a long period of time because the current frame is dependent on previous frames. This string of errors reduces the quality of the radio user’s voice. When the vocoder uses a prediction coefficient (between 0 and 1) to construct the parameters of the current frame, the dependence on previous frames is lessened. DVSI has determined that prediction coefficients of 0.7 for low-pitched speakers and 0.4 for high-pitched speakers produce the best coding gain with the lowest error persistence.
Interleaving is another process that reduces the potential of channel errors affecting a large section of the reproduced speech. When errors occur in bursts – a frequent occurrence as signal power fades – they can break error control codes and create a noticeable gap in the audio at the receiving radio.
Interleaving combats this problem by actually “mixing up” the bits as they are transmitted over the air. With this method, an error burst will be broken into several small error pieces that are more easily corrected by the error correcting codes.
Channel degradation is lessened by using various means of error mitigation to reduce the impact of uncorrected errors. These strategies include adaptive smoothing, which reduces the frame-to-frame variation in frames with high error rates, and frame repeating, or repeating the parameters of the previous frame when a frame contains a high number of estimated errors. These error mitigation strategies are important to the user because the radio output will be muted if the prevalence of errors becomes too great for the vocoder to accurately represent the speech.
Vocoder technology also can be enhanced through forward-error correction, a process in which additional bits are transmitted along with the speech bits used to detect and correct errors that occur during the transmission.
Taken together, these design elements and enhancements allow vocoders to present close representation of a radio user’s voice. Test results from the TIA-funded study would seem to bear this out. Radio users were asked to listen to vocoded speech over four vocoders operating under a variety of conditions and to rate the speech on a scale of one to five, with five being the highest. The radios were used while stationary and while at different speeds to achieve fading characteristics. Under each of the seven operating conditions, the IMBE vocoder scored highest.
Improved vocoder technology is expected to gain rapid acceptance as a part of digital trunked radio. Ericsson is using the IMBE vocoder in many of its radios and mating the technology to its Enhanced Digital Access Communications System (EDACS) trunked radio system.
Adoption of improved vocoders such as the IMBE in mobile and portable radios has proven in studies to be an effective method of enhancing the quality of digital trunked radio systems. Combined with the other benefits of digital trunked radio-fast access, high capacity and security-the IMBE vocoder will give law enforcement agencies and other users many reasons to consider going digital.
Dominick Arcuri, product manager for EDACS Systems at Ericsson Private Radio Systems in Lynchburg, VA. IMBE is a trademark of Digital Voice Systems. EDACS is a trademark of Ericsson Private Radio Systems
