The QoS Solution: Fix Circuit-Switched Voice Codecs in a Packet-Switched, Wireless World with Enhanced Speech-Processing Software

If circuit-switching voice codecs are the challenge to good QoS in wireless, packet-switched networks, what then is the fix for outdated voice codecs? A market of enhanced speech-processing software is emerging that corrects for the shortcomings of traditional voice codecs that were designed decades ago for a circuit-switched PSTN. These recent developments in Vo802.11b software provide QoS enhancement solutions for IP telephony in the terminal with very high voice quality even with severe network degradations caused by jitter and packet loss. These Vo802.11b QoS enhancements should provide Vo802.11b speech quality comparable to that of the PSTN. Also, speech quality should degrade gradually as packet loss increases. Moderate packet loss percentages should be inaudible.

Enhanced Speech-Processing Software

New speech-processing algorithms provide for diversity, which means that an entire speech segment is not lost when a single packet is lost. Diversity is achieved by reorganizing the representation of the speech signal. Diversity does not add redundancy or send the same information twice. Ergo, it is bandwidth efficient and ensures that packet losses lead to a gradual and imperceptible degradation of voice quality. The tradeoff is that diversity leads to increased delays. Enhanced speech-processing software includes advanced signal processing to dynamically minimize delay. Therefore, the overall delay is maintained at approximately the same level as it would be without diversity. Furthermore, the basic quality (no packet loss) is equivalent to or better than PSTN (using G.711).

Enhanced speech-processing software is built to enhance existing standards used in IP telephony. This software enables high speech quality on a loaded network with jitter, high packet losses, and delays. Cost savings are realized using enhanced speech-processing software, as no need exists to overprovision network infrastructure. The high packet loss tolerance also reduces the need for and subsequent cost of network supervision, resulting in further cost savings.

Examples of Enhanced Speech-Processing Products

Voice Quality in Vo802.11 can be made to equate that of the PSTN by implementing a number of measures. The following pages will explain those measures and how they improve voice quality.

Adaptive Jitter Buffer The use of an adaptive jitter buffer optimizes sound quality by using an advanced adaptive jitter buffer control combined with an error concealment algorithm. One such product is GIPS NetEQ™ from San Francisco-based Global IP Sound. It works with any codec such as iLBC (GIPS low bit rate codec), G.711 (including GIPS Enhanced G.711), G.729, and G.723.1. NetEQ is only required on the receiving end of any conversation. NetEQ improves the sound quality significantly without any interoperability problems. This solution quickly adapts to the dynamic network conditions of packet-switched networks. This ensures high speech quality with significant latency savings compared to conventional jitter-buffering technology.

Enhanced G.711 G.711 with enhancement provides superior packet loss robustness. Enhanced G.711 consists of the G.711 codec combined with an enhancement to provide packet loss robustness. During the call setup, it is determined if the recipient also has Enhanced G.711. If so, the call will continue using GIPS Enhanced G.711; if there is not a match, the call will proceed using G.711 on both ends. The enhancement unit is similar in function to encryption methods. The packets are transcoded to prevent packet loss as opposed to privacy. Enhanced G.711 in combination with an adaptive jitter buffer provides a PSTN speech-quality level at packet loss/delay rates up to 30 percent. This is achieved without increasing the bit rate and without increasing latency significantly.

Packet Loss Robustness Using a low bit rate codec is one method for increasing packet loss robustness; low bit rate codecs use less bandwidth, providing a more efficient use of the available bandwidth. The basic speech quality of one low bit rate codec, GIPS iLBC freeware, offers better voice quality than G.729 and G.723.1, and it operates at a rate of 13.3 Kbps. Another method of increasing robustness to packet loss is to use an error concealment algorithm, such as the aforementioned GIPS NetEQ.

Acoustic Echo Cancellation Echo is often prevalent when using a PC or IP phone. Acoustic echo cancellation is contained in enhanced speech software to counter echo in those platforms.

Results of Enhanced Speech Software: An Independent Evaluation by Lockheed Martin

Lockheed Martin Global Telecommunications (LMGT) recently performed independent subjective tests of speech enhancement software products from San Francisco-based Global IP Sound. The tests show that GIPS speech and audio-processing software NetEQ, GIPS Enhanced G.711, and iPCM-wb outperform the legacy VoIP enhancement products in their respective categories. When installed in the IP edge devices, such as media gateways and IP phones, GIPS solutions give equipment and network service providers outstanding QoS even during peak traffic hours.

The independent test laboratory within LMGT (previously a part of COMSAT Laboratories) conducted the subjective testing of the GIPS sound-processing software. The tests compared conventional speech codecs with GIPS NetEQ, adaptive jitter buffer and error concealment, GIPS Enhanced G.711, a standard-improvement solution, and iPCM-wb, a high-end (wideband) speech processing codec. The LMGT Test Report concluded, "The perceived speech quality performances of the GIPS telecommunications and wideband solutions show a dramatic improvement of that of traditional G.711 for the range of impaired-channel scenarios tested."

The test results reveal that even under situations with high packet loss and delay, Global IP Sound's enhanced speech-processing software for telephony bandwidth delivers a sound quality that matches or is very close to PSTN quality. The GIPS wideband speech solutions for end-to-end IP applications deliver higher than PSTN quality with very high robustness to network degradations.

The test results verify that NetEQ, which is codec independent, significantly improves the sound quality of the standard G.711 codec under conditions with packet loss/delay. The combination of GIPS Enhanced G.711 and NetEQ delivers sound quality matching PSTN for moderate packet loss/delay levels while being very close to PSTN for high packet loss/delay. The combination of iPCM-wb and NetEQ-wb (products aimed at the broadband market) offers high-fidelity sound quality even under adverse network conditions with up to 30 percent packet loss/delay (see Figure 6-7).

Figure 6-7: Results of Lockheed Martin testing of voice quality using Global IP Sound-enhanced speech software products. Note that a voice quality comparable to the PSTN is achieved. Source— Global IP Sound

The methodologies and test listening groups were selected in accordance with acknowledged criteria set out by the ITU, ETSI, and the TIA, the relevant international standards bodies.^[13] The conclusion of these tests is that, given the vagaries of both VoIP and wireless environments, enhanced speech software can correct most adverse conditions in Vo802.11 regarding latency, jitter, and packet loss and still deliver a voice quality that is comparable to that of the PSTN.

Objection 2: Security for Vo802.11

Although a chapter was devoted earlier in this book to security on 802.11 networks, it is important to examine how security applies specifically to Vo802.11. The misperception is that, because the conversation is transmitted over the airwaves, the voice stream is susceptible to interception, that is, eavesdropping. Although such an occurrence is not entirely impossible, it would be extremely difficult to tap into such a conversation.

Case Study for Security of Vo802.11: SpectraLink's Secure Radio Technology Vo802.11 telephone systems provide additional measures of security through sophisticated radio technology and proprietary signal encoding. Many Fortune 500 companies use Vo802.11 telephone systems in their most secure areas, such as executive offices, data centers, and network control centers. Vo802.11 telephone systems employ digital spread spectrum transmission and a pseudorandom hopping sequence against radio eavesdropping.

Digital Spread Spectrum Transmission The Vo802.11 telephone systems use a proprietary implementation of frequency hopping spread spectrum radio transmission, a radio technology originally developed by the military for secure and covert communications. Spread spectrum takes a discrete signal, such as a digitized voice conversation, and spreads it over a wide range of frequencies rather than transmitting at a single carrier frequency. Vo802.11 phone systems use frequency hopping to spread the signal by changing the carrier frequency once every 10 milliseconds (100 times every second). Because the carrier frequency changes rapidly, a radio scanner or narrowband receiver cannot be used to recover the information. These systems use 25 to 50 different frequencies in the hopping sequence, so a narrowband scanner has access to less than 4 percent of any conversation.

Proprietary Pseudorandom Hopping Sequence A proprietary pseudorandom hopping sequence provides additional security. If a potential eavesdropper went to the expense of developing a scanner that could change frequencies 100 times a second, he or she would then have to attempt to determine the pseudorandom sequence to know when to monitor which frequency. Furthermore, each base station is transmitting on a different frequency, so multiple scanners would be required to follow a conversation as it was handed off from base station to base station.

Vo802.11 systems use digital transmission, meaning that the analog voice signal is converted to a digital signal. This digital signal is scrambled to improve transmission, further complicating the ability to interpret an intercepted signal. Finally, Vo802.11 systems use Time Division Multiple Access (TDMA) to provide multiple speech channels from a single base station. The frame format and signaling for a Vo802.11 TDMA signal are proprietary and would have to be determined to identify discrete conversations on the radio link.^[14]

Objection 3: CALEA and E911

An objection to VoIP often posed by the circuit-switched industry concerns a telecommunications carrier's compliance with CALEA and E911, which are legal requirements for primary-line telephone service providers in the United States. The laws requiring telephone companies to provide these services were made before the Internet became part of mainstream America. Although the technological means for providing these services in a number of circumstances exists, the first question that should be asked is what obligation does the service provider have in providing these services?

802.11 is, at the time of this writing, primarily used in enterprise environments. An enterprise has no obligation to provide itself with CALEA or E911 services. Where these requirements may gain greater scrutiny is in residential applications along the lines of providing "lifeline services." 802.11 Internet access is provided by wireless Internet service providers (WISPs). The Federal Communications Commission (FCC) considers ISPs (and WISPs) to be information service providers, not telecommunications service providers. The distinction between them is that information service providers are not required to provide the same scope of services required of telecommunication service providers. Therefore, a WISP is under no obligation to provide CALEA or E911 services.

WISPs are also considered to be second-line service providers, as opposed to primary-line service providers. Primary-line service providers are considered telecommunications service providers and are expected to comply with the same laws that incumbent telephone companies must comply with. A short list of those requirements includes CALEA, E911, paying into the Universal Service Fund, and paying access fees when originating or terminating longdistance calls.

Finally, a WISP may have little or no knowledge of its subscribers using Vo802.11. No unique equipment or recording is required of the WISP for voice calls to originate or terminate on its network. All a WISP does is provide access to an IP network (namely the Internet or a managed IP network). The subscribers then equip themselves with IP handsets, Vo802.11 phones, VoIP software on their PCs, and so on. In such a scenario, the subscriber is not required to provide themselves with CALEA or E911.

E911 A number of E911 solutions are arriving on the market at the time of this writing. First, some solutions are overflowing from the cell phone industry where E911 will soon be a requirement. One solution in the case of Vo802.11 is a mechanism to triangulate the signals emanating from the victim's 802.11 handset. Different access points can report the vector and distance to the subject Vo802.11 handset. Another solution is to include Global Positioning Satellite (GPS) technology in a Vo802.11 handset. That way the exact location of the handset is known at any time.

CALEA This requirement may be relaxed in a forthcoming regulatory regime outlined by FCC Chairman Michael Powell in an October 30, 2002 address at University of Colorado. In that speech, Chairman Powell conceded that the CALEA law was designed for the circuit-switched world and was, at the time of that speech, almost impossible to comply with in a Vo802.11 environment. As a result, and in the interest of promoting all that 802.11 and similar technologies have to offer, Chairman Powell hinted that such requirements would have to be relaxed.

It should also be emphasized that CALEA and E911 are requirements for the U.S. market. Vendors would do well to not focus solely on the U.S. market and let these requirements for the circuit-switched world restrict them in developing products and services that would benefit consumers around the world.

^[13]Third-party Evaluation of GLOBAL IP SOUND Edge Device QoS Solutions for VoIP, a white paper from Global IP Sound, available online at www.globalipsound.com.

^[14]Geri Mitchell, "Radio Security of the Link Wireless Telephone System," a white paper from SpectraLink, www.spectralink.com.