Better Speech Recognition with Digital RF System in Study of Cochlear Implants
Jace Wolfe, Mila Morais, Erin C. Schafer, Emily Mills, Robert Peters, Leslie Lianos, Andrew B. John, Mary Hudson · The Hearing Journal · 2013
Hearing aid wearers often have difficulty understanding speech in the presence of background noise and reverberation (J Acous Soc Am 2008;124[5]:3064-3075; J Acoust Soc Am 1978;63[2]:533-549). They also may run into trouble when the signal of interest originates from across a large room. Remote microphone hearing assistive technology (HAT), more commonly referred to as a personal frequency modulation (FM) system, generally is the most effective method to improve performance in such challenging listening situations (J Speech Hear Disord 1984;49[4]:409-418; J Speech Hear Disord 1984;49[3]:278-286; Am J Audiol 2010;19[1]:36-45). These systems capture the signal of interest near its source, leading to a much more favorable signal-to-noise ratio (SNR) compared with that at the microphone of the hearing aid. Taking this information into account, we compared different types of hearing assistive technology to see if one approach was better than another in these difficult listening environments. We found that cochlear implant recipients using a digital radio frequency (RF) system had better speech recognition, especially at high noise levels, than those using other technologies. FREQUENCY-MODULATED RADIO WAVES Personal FM systems deliver the signal of interest from a transmitter to a receiver via a radio frequency signal. In FM transmission, the desired signal is modulated in the frequency domain and overlaid on a sinusoidal high-frequency carrier wave. The frequency on your car radio dial corresponds to the carrier frequency. By contrast, AM signals are modulated in the amplitude domain before being overlaid on the carrier wave. In both types of transmission, the receiving device—for example, your antenna and car radio—captures and demodulates the transmitted signal, in effect separating the desired signal, such as music, from the carrier. While a full discussion of the difference between AM and FM transmission is beyond the scope of this article, it is useful to note a few of the reasons why FM is a more desirable transmission method for hearing assistive technology. First, FM signals have better sound quality than AM signals. This is because the Federal Communications Commission has protected higher frequency ranges for FM transmission, providing wide bandwidth for transmission and the potential to use multiple FM channels. Second, FM signals are less prone to interference than are AM signals. Consider the difference in quality and the amount of static audible on your car radio's AM and FM stations. Finally, because FM waves are broadcast at a higher carrier frequency than AM waves are, FM waves have a lower ability to pass through physical barriers. This means that FM signals are less prone to “leaking” from one room to another. However, they're certainly not immune to this problem, which will be discussed later. DIGITAL RF TRANSMISSION The benefits of digital signal processing are well established. Most importantly, in hearing technology applications, digital processing allows for more complex and precise analysis and manipulation of a signal compared with analog processing.Figure 1: This analog signal is converted to a digital signal with a three-bit system.Analog signals are converted to digital signals through sampling of the original signal of interest. Figure 1 shows an analog signal that is sampled at different points in time. The amplitude of the signal at these points is coded through discretization and quantization into “bits,” a contraction of the words “binary digits.” A bit of digital information is represented as one of two digits—0 or 1. Let's say that we would like to code the signal in Figure 1 with a three-bit system. The relevant amplitude range of the signal of interest is divided into eight values (three opportunities to classify the signal, with two options each [0 or 1]; 23 = 8). First, the digitizer determines whether the amplitude of the signal falls within the top or bottom half of the relevant range of amplitudes. If it falls within the top half, a “1” is assigned, and if it falls within the bottom half, a “0” is assigned. The process is repeated twice more, with the top and bottom halves again split into two and a “1” or “0” assigned depending on where the signal resides.Figure 2: In amplitude-shift keying, a “0” is depicted by a cessation in the carrier wave, and a “1” is depicted by the presence of the carrier wave.Amplitude-shift keying (ASK), as illustrated in figure 2, is one approach used for digital RF transmission. Just as with AM and FM radio, a carrier frequency delivers the signal of interest. However, the carrier frequency is modulated between “on” and “off” states by a binary digital code. When the system must code a “1,” the carrier frequency is transmitted at the specified amplitude. Conversely, when the system must code a “0,” the carrier frequency is temporarily muted.Figure 3: In Gaussian frequency-shift keying, a positive frequency deviation is used to represent a “1” and a negative frequency deviation a “0.”Gaussian frequency-shift keying (GFSK) is another method for digital RF transmission. In this approach, a “1” is represented by a positive frequency deviation imposed on the carrier frequency, and a “0” by a negative frequency deviation imposed on the carrier frequency (seefigure 3). A potential advantage of GFSK over ASK is less susceptibility to noise/interference. In most contemporary digital RF systems, the carrier frequency is much higher than that used in AM and FM radio applications. In Bluetooth technology and the system described in this article, the carrier frequency is 2.4 GHz. Because many commercial devices feature digital RF transmission, the carrier frequency continually “hops” from one carrier frequency to another 800 times or more per second, dramatically reducing the potential for interference between devices close together. During the roving process, many systems will identify and avoid channels in which interference may exist. THEORETICAL ADVANTAGES The advantages of digital radio frequency are many. First, by means of the control data broadcast along with the digitized audio signal, analysis and manipulation of the transmitted signal is much more precise and sophisticated compared with analog systems. As a result, it may be possible to be more effective in enhancing the speech signal of interest and reducing background noise. Second, as every educational audiologist is fully aware, interference may occur when two children with personal FM systems use their devices within the same building. Interference may manifest in the form of static noise, or one child may hear the voice of the other child's teacher rather than her own teacher. With a digital radio frequency system such as Phonak Roger, many systems can coexist in the same vicinity without interference. Each Roger system repeatedly broadcasts audio packets at different RF frequencies. If the first packet is lost, the second will be heard. If that one is lost, a third one may be heard, and so on. The creation of networks and subnetworks is easy, and channel planning or channel allocation is not required. Third, digital RF allows for transmission of a wider bandwidth of the signal of interest. This may be beneficial given recent research indicating the importance of information above 6,000 Hz for children with hearing loss (J Speech Lang Hear Res 2008;51[3]:785-797; Ear Hear2002;23[4)]316-324; J Am Acad Audiol 2010;21[10]:618-628). ROGER DIGITAL RF Phonak Roger is a digital wireless radio system allowing for long-range, low-delay, full audio broadcast to miniature low-power receivers. The system operates in the 2.4 GHz industry, science, and medical band, and it uses the GFSK approach to deliver the digital RF signal. Roger is globally license free, uses adaptive frequency hopping to avoid electromagnetic interference, and has an audio bandwidth around 7,300 Hz. Digital processing is paired with a new algorithm that controls adaptive gain changes, with the intention of improving the signal-to-noise ratio compared with traditional adaptive FM systems, also known as Dynamic FM systems, especially in high noise conditions. Performance with the Roger digital RF system was evaluated in seven adult cochlear implant recipients who had used their implants for six months or longer and scored at least a 50 percent for correct responses on the consonant-nucleus-consonant (CNC) monosyllabic word recognition assessment in quiet. All participants used OPUS 2 sound processors with MED-EL SONATA TI100 cochlear implants. The design of this research project has been described extensively (J Am Acad Audiol 2009;20[7]:409-421). To summarize, speech recognition in noise was assessed in a classroom-type environment (22' x 4” in length; 15' x 5” in width; 8' x 9” in height). Hearing in Noise Test (HINT; J Acous Soc Am 1994;95[2]:1085-1099) sentences were presented from a loudspeaker located in the front of the classroom, and recorded classroom noise (Am J Audiol 2006;15[2]:114-126) was presented from four loudspeakers located in the corners of the classroom. The level of the HINT sentences was 85 dBA at the microphone of the RF HAT transmitter and 64 dBA at the location of the subject. Sentence recognition was measured in quiet and at each of the following competing noise levels—50, 55, 60, 65, 70, 75, and 80 dBA—in three different conditions: Cochlear implant coupled to Phonak Roger digital RF system. Cochlear implant coupled to Phonak MLxi Dynamic FM system. Cochlear implant coupled to Phonak MLxS fixed-gain FM system. The Phonak MLxS FM system was programmed to provide a +10-dB FM advantage for the signal coming from the FM receiver relative to the signal from the microphone of the sound processor. This FM advantage is fixed regardless of the ambient noise level. The Phonak MLxi Dynamic FM system changes the gain of the FM receiver in order to provide a more favorable FM advantage in noisy environments. When no speech is presented to the microphone of the FM transmitter, the FM advantage of the system is set to -16 dB, which in practice means that the receiver is muted. When speech is presented to the FM microphone and the ambient noise level is below 57 dB SPL, the FM advantage of the system is set to +10 dB. The FM advantage increases adaptively once the ambient noise level exceeds 57 dB SPL, until it reaches a maximum FM advantage of +24 dB. The subjects and examiner were blinded to the randomized RF HAT system used in each condition. BETTER SPEECH RECOGNITIONFigure 4: Mean speech recognition in noise was measured as a function of competing noise level for seven MED-EL cochlear implant users with three different radio frequency systems.Figure 4 shows the results of the study. A repeated measures analysis of variance and post-hoc analyses with the Tukey–Kramer multiple comparisons test indicated a significant main effect of device (F [2, 168] = 20.5, p = 0.0001) and signal level (F [7, 168] = 45.5, p .05). The post-hoc analysis of the two-way interaction effect between device and signal level suggested that, at the 75 and 80 dBA noise conditions, subjects performed significantly better with the Roger digital RF system than with the two FM receivers (p <.05). Additionally, the Phonak MLxi Dynamic FM system provided significantly better speech recognition in the 75 dBA noise condition than the Phonak MLxS fixed-gain FM system did (p <.05). The results of this study indicate that MED-EL cochlear implant users will potentially achieve better speech recognition in noise with the Phonak Roger digital RF system than with dynamic or traditional FM systems, particularly at high noise levels. The benefits observed with the Roger digital system are similar to benefits experienced by Advanced Bionics and Cochlear users in a similar study (Wolfe et al: J Am Acad Audiol; in press). Linda Thibodeau, PhD, also measured better performance in noise with the Roger digital RF system compared with dynamic and traditional FM for a group of hearing aid users (Phonak's Advances in Audiology conference; Dec. 2-5, 2012; Las Vegas). Although it was not formally assessed in this study, an additional, and quite valuable, real-world benefit of the Roger digital RF system is the elimination of the need for the audiologist to manage RF transmission channels, which is required with other systems to avoid the potential for interference in educational and social settings. None of the subjects reported any negative experiences associated with use of the Roger system. ACKNOWLEDGEMENTS The study was funded by a grant from Phonak. The authors thank Felix Goldbeck, Hans E. Mülder, and Francois Marquis of Phonak, who served as consultants during the study. References on Tap Access the hyperlinks (shown in gray) in this article and throughout the issue by reading it on HJ's free iPad app: http://bit.ly/AppHearingJ .