Localization 102: Essentials of a Home-based Front-back Localization Training Program
Tech Topic | January 2015 Hearing Review
It is critically important to provide all the natural localization cues through the hearing aids. This study shows that localization performance can be enhanced through the use of listener-based training programs.
In a previous paper,1 we described how current hearing aids can preserve acoustic cues that are important for localization. We also indicated how these cues may not be immediately useful to the hearing-impaired listeners whose auditory system may have maladapted to receiving (and interpreting) such cues.
The use of training programs may be necessary to re-train the auditory system to utilize the new cues to yield optimal outcome. This paper describes the considerations behind the development of the home version of the Widex ORCA Localization Training (WOLT) program and the results of a study evaluating the program. An expanded version of the study can be found in Kuk et al.2
Why Develop a Localization Training Program for Hearing Aid Wearers?
It was previously shown that localization accuracy is impaired in listeners with a hearing loss.1 Average accuracy was only between 30% and 45% when identifying sounds from the front/left/right, and around 20% in identifying sounds from the back. Thus, in hearing-impaired listeners, front/back localization is more impaired than front/left/right localization. This could raise significant safety concerns and confusion, as well as compromise the quality of life of listeners with a hearing impairment.
Providing the right acoustic cues is only part of the solution. Kuk et al3 showed that providing hearing aid wearers with the pinna compensation cues alone improved their localization ability. However, listeners who were trained with the interaural and pinna compensation cues achieved a higher localization performance than just given the cues alone.
Kuk and Keenan4 also showed in listeners with an unaidable high-frequency hearing loss that training plus the use of the Audibility Extender (a frequency lowering algorithm from Widex) improved their speech understanding scores more than either one of these approaches alone. Other investigators have also reported on the efficacy of training in enhancing the outcomes of hearing aid interventions. Sweetow and Palmer5 provided a review on the topic.
Despite the fact that there is clearly room for improvement on localization ability and the proven efficacy of training, there is a paucity of training programs that are geared toward localization training. There are references to localization training in the cochlear implant literature (eg, Preece, 20106); however, these training programs are primarily targeted toward left/ right discrimination (eg, Tyler et al, 20107).
In addition, almost all of these training programs are structured in a hemi-field (ie, half-circle or 180°) where an assessment of front/back localization is impossible. This is a limitation considering that discrimination between front and back is a more common difficulty than left/right discrimination that hearing aid wearers encounter. This prompted us to design a localization training program where the focus is on front/back localization.
Considerations in Developing the Training Program
There are many rehabilitation training programs (eg, LACE8) with reported efficacy and effectiveness. A common factor that determines their acceptance by their targeted audience (in this case, hearing-impaired listeners) is whether they keep the users’ motivation. Listeners who are motivated are more likely to show an improvement of their performance.
Thus, we paid special attention to factors that would maintain the users’ motivation when designing the WOLT program. Here are some of the considerations:
- Be short and sweet. One of the outcomes of learning studies is that training spread over a longer period of time leads to more improvement than focused training (same number of hours packed into a shorter time frame).9 Thus, training for a short period of time regularly is encouraged.
- Be practical. To encourage frequent training, the training program has to be available in a format that is easily accessible from the listeners’ home using equipment that the average household has.
- Focus on relevant parameters. It is fair to assume that most households have a computer that can be used in training. However, it is not likely that the average household has loudspeaker arrays that are typically used in localization studies. Given that one is restricted to using only two loudspeakers, one has to be realistic and practical on what can be trained. Because hearing-impaired people have the greatest difficulty discriminating between front and back (and not as much on right and left), one should focus on front/back training.
- Make the task simple. It is normal for listeners to remain motivated on a task if they know what to do. Thus, keeping the instructions and the task (or what is required of the listeners) simple is critical to ensuring motivation. Asking listeners to point to the front or back in response to the stimulus is acceptable.
- Provide feedback (learning opportunities). The training program should result in listener learning through the experience. The listeners should learn what “front” sounds like and what “back” sounds like. This stimulus-response association strengthens one’s learning. Thus, the listeners should repeatedly listen to the stimulus and its intended correct response to strengthen the association. Several researchers have included the provision of feedback in their training programs.7,10
- Make it meaningful. In order to improve the listeners’ ability to localize sounds in their daily environments, it would be appropriate to use daily environmental sounds as stimuli during the training. It makes the stimuli relevant to the listeners and promotes generalization to other stimuli. For example, Tyler et al7 used 16 different everyday sounds in their training.
- Make the training adaptive. One of the key principles to facilitate learning is to present materials that are at the right level for the learner.10 The materials cannot be too easy because the learner will find them boring and refuse to participate; likewise, it cannot be too difficult because the learner may lose confidence and interest (because s/he always fails) on the task. Finding the happy medium and keeping them motivated is a key to successful learning. This is a reason why many training programs today are “adaptive” in nature so that the stimuli are always challenging but not impossible to the learners. In training localization, one may start by using stimuli that are easier to localize first. As the listeners achieve a certain performance level, the stimuli can be made more difficult. The following parameters are known to affect stimulus difficulty in localization.
- Stimulus duration. Longer stimuli are easier to localize than shorter stimuli.11
- Front/back difference. The spectral difference between a stimulus presented from the front and from the back is used as a cue for front/back discrimination. Our pilot study suggested that, if this difference is made larger than what the pinna shadow provides (2 dB on HR_20_25_Kukks-USE.indd 21 average above 2000 Hz), it would make front/back discrimination easier.
Figure 1 shows the unaided localization scores for five different stimuli in each of the front/back/left/right directions collected in 15 subjects. These stimuli differ in their duration and back attenuation. They include 300 ms without back attenuation, 500 ms without back attenuation, 1s with 2 dB back attenuation, 2s with 4 dB back attenuation, and 3s with 8 dB back attenuation. It is seen that stimulus duration and/or back attenuation does not affect the accuracy of localization (around 60%) for the front, left, and right directions. However, stimulus duration/back attenuation affects the accuracy of localization for sounds originating from the back. The longer and the more attenuated the stimulus, the higher the back localization scores.
Thus, if one were training front/back localization, one may need to artificially increase the intensity difference between stimuli presented from the front and the back at the beginning of the training. As the listener gains competence, one may gradually decrease the intensity difference between the front and back stimuli to a more natural level.
Description of the WOLT Program
The home training program includes a custom software program written in Visual Basics that can be loaded into any personal computer with an operating system higher than Windows XP. It requires that the computer has a standard 2-channel sound card and two external loudspeakers. They can be any portable loudspeakers that are easily connected to the computer. In our studies, we used the Logitech S-20 loudspeakers. The loudspeakers should be matched in electroacoustic characteristics (frequency response and output limit). The cable connecting the two loudspeakers should be long enough to allow the loudspeakers to be separated by at least an arm’s length.
Stimuli included in the home training program are 30 seconds chosen from the Life Sounds Library (from Widex Compass software) and from the album 500 Ultimate Sound Effects by Dr. Sound Effects (published by HDsoundFX 2010). The soundtracks include household, office, city, human, nature, hardware, sports, impact, emergency, and animal sounds. Only sounds that are judged to be familiar by the staff at ORCA were selected. These sounds include speech, music, animal sounds (eg, dog bark, cow moo), environmental sounds (church bells, train, police siren, etc), and home sounds (telephone ring, alarm clock, silverware clanging, etc).
Each sound is high-pass filtered at 2000 Hz to ensure that only high frequency cues, which are critical in front/back localization, are retained and are available for localization training. The level of the stimuli is matched to the peak RMS. Each sound is available in 5 durations (3000 ms, 2000 ms, 1000 ms, 500 ms, and 300 ms) by 4 front/back attenuation (8 dB, 4 dB, 2 dB, and 0 dB) settings for adaptive difficulty. To increase efficiency, the 5 x 4 or 20 combinations of stimulus durations and attenuations are reduced to only 5 combinations: 3000 ms with 8 dB attenuation, 2000 ms with 4 dB attenuation, 1000 ms with 2 dB attenuation, 500 ms with 0 dB attenuation, and 300 ms with 0 dB attenuation. The easier stimulus condition (long duration and large back attenuation) is presented first; the more difficult stimulus conditions are introduced when the listeners reach a criterion level of performance. Only 10 random sounds are used during each training trial. Each sound is presented 3 times through the front and back loudspeakers in a random order.
To prepare the listeners for the training, they need to be instructed on how to position each loudspeaker so they are seated at equidistant to the front and to the back loudspeakers. The clinician should demonstrate the setup, train with the listeners at the clinic until they understand the task, and have the listeners repeat the setup satisfactorily before they return home for training.
During the training, the listeners would need to first adjust the overall level of the output so all sound samples (which are calibrated to have the same peak rms value) are comfortably loud. Figure 2 shows the screen display where the listener can adjust the overall presentation level.
During stimulus presentation, the listeners are required to indicate if the sound comes from the front or the back loudspeaker by pressing the appropriate button on the computer screen. A correct response turns the button green. An incorrect response turns it red, while at the same time, the button for the correct location turns green (Figure 3).
The listeners have the opportunity to compare the incorrect location to the correct location by clicking each button as many times as they wish. They click the “NEXT” button when they want to hear the next test stimulus.
At the end of a training block, the listeners see their results displayed along with results from previous training blocks (up to 10 blocks, Figure 4). In order not to confuse or discourage the listeners with what criteria or stimulus conditions are used (but still provide feedback), a color scheme is used to represent gross performance. The warmer colors (like red) represent poorer performance while the cooler colors (like green) represent better performance. However, the listeners are not informed of the test conditions. This way, the listeners can be informed of their performance without being bogged down by details. The listeners are asked to complete as many training blocks as they can for 30 minutes per day. This can be in one sitting or split into several smaller trials, 5 times per week. A total of 4 weeks of home training are recommended.
The durations and back attenuation values of the home training stimuli change adaptively based on the listener’s performance. All listeners start training with the stimuli at the easiest level of 3000 ms duration and 8 dB attenuation. Each training level is tested twice and the back scores are averaged to determine the stimulus condition for the next block of tests. To avoid biasing the back loudspeaker, the localization scores for the front loudspeaker must be >50% in order for the listeners to advance to the next level. The data from the home training can be saved to a USB memory stick that can be downloaded by the clinician for examination of progress when the listeners return for service.
Field Study: Outcome of Home Training Program
We evaluated the efficacy of the training program on 10 hearing-impaired listeners. All 10 were fitted with the wireless Widex Clear m-CB hearing aids in the bilateral mode at the default settings and with the IE-compression (for interaural level cues) and DP (for pinna compensation). The m-CB model is a BTE with a reported in situ bandwidth as high as 10,000 Hz, which could be important for the back and vertical localization. Thus, all listeners were provided with the necessary acoustic cues for localization. All listeners wore the hearing aids for a month prior to the study.
The listeners’ baseline localization ability was evaluated with 3 stimuli, all 300 ms in duration (no attenuation of sounds from the back, as in training), prior to the training. These stimuli included a narrow-band noise, a female speech passage, and a telephone ring. All the stimuli were filtered above 2000 Hz. Stimuli were presented from 12 loudspeakers arranged in a full circle with a separation of 30° between loudspeakers. Afterwards, 5 of the listeners wore the hearing aids home without any special instructions on localization (control group). The other 5 listeners were trained on the home localization program for 1 month (study group). All returned after 1 month to determine their localization ability with the same test stimuli.
Figure 5a-b. Hypothetical CoM representation of: a) perfect localization (red line matches outer circle), and b) poor localization where all responses are indicated as originating from the front and sides.
The Center of Mass (CoM, represented by the solid red line) approach is used to display the localization data. This method has been described in detail in other studies on sound localization.11,12 A simple way to interpret the localization data represented by the CoM is to examine how closely the CoM (red line shown in Figure 5) is toward the outer circle. The closer the CoM is toward the perimeter of the circle, the more accurate the localization.
Figure 5a shows an illustrative pattern where the listener has perfect localization to stimuli from all azimuths. Figure 5b shows the response of a listener where all the responses are indicated as originating from the front and sides.
Control group with no training. The CoM data of the control group at the beginning of the study with the study hearing aids (Figure 6a) and 1 month into the study (Figure 6b) are shown in Figure 6.
The average listener in the control group had significant front/back localization difficulty. The responses to stimuli presented from the back were mostly indicated as originating from the front.
Figure 6b also shows that there was not much change in the response pattern after the listeners had used the hearing aids for the 1-month study period. One should be reminded that all listeners have available pinna cues and interaural level cues from the use of Clear hearing aids 1 month prior to the study. Any effect of the acoustic cues (such as the inter-ear and DP) may have already been realized before the study started. Thus, the lack of a difference suggests no additional improvement without intervention.
Effect of training for study group. Figure 7 shows the CoM data of the study group. Prior to the training, listeners in this group performed as poorly as the control group listeners. Most of the errors were front/back errors.
With training, Figure 7b shows that there was a significant expansion of the “red circle” toward the perimeter of the circle.
This suggests better localization in all directions including the front and back after 4 weeks of home training. The improvement in localization is statistically significant (p < 0.05). It is important to remember that only front/back training was provided during the training period.
One of the criticisms in studies of localization training is that generalization seldom occurs or that it occurs at a much slower rate than the stimuli that are used for the training.13 It is of interest to note that the training stimuli used in the home training program are entirely different from the stimuli that are used in the testing. The data on the “before-after” testing of localization provide proof that generalization occurs with home training using multiple everyday stimuli.
Previous studies at the Widex ORCA laboratory showed that the correct cues (ILD and pinna cues3) improve localization abilities. This study shows that localization ability can be further improved with training. Improvement is not restricted to only the stimuli used to train localization, but also for sounds that are not used during training. In addition, even though the training is conducted on front/back discrimination, generalization to other directions is also possible.
We believe that the ability of the training program to maintain the motivation level of the listener is vital to the success of the program. In our training, we start with an easier stimulus (longer duration and greater front/back contrast) and adaptively adjust its difficulty based on the listeners’ performance. In order to encourage self-correction in the next trials, participants are given visual and auditory feedback on their performance during the training (listen and compare) and after the training so they know which directions require more attention. To minimize frustration that may decrease motivation, we inform listeners of their progress using a color scheme while bypassing all the unnecessary details. The intended result is that the listeners feel they have made improvements on every trial, and that they are motivated to achieve better performance in the next trial. All these elements are important in motivating the listeners.
In summary, we believe that it is critically important to provide all the natural localization cues through the hearing aids. This study shows that localization performance can be further enhanced through the use of listener-based training programs. In selecting a training program for the listener, it is important to consider what is being trained, and if the algorithm used in the training program keeps the listener’s interest in learning. With the appropriate program, it is possible that significant improvement can be measured in as little as 1 month’s training. (Authors’ Note: If you are interested to obtain a free copy of the training program for your patients, e-mail email@example.com for a link and suggestions on how to use the program.)
- Kuk F, Korhonen P. Localization 101: Hearing aid factors in localization. Hearing Review. 2014;21(9):26-33. Available at: http://www.hearingreview.com/2014/08/ localization-101-hearing-aid-factors-localization
- Kuk F, Keenan D, Lau C, Crose B, Schumacher J. Evaluation of a localization training program for hearing impaired listeners. Ear Hear. 2014; In press.
- Kuk F, Korhonen P, Lau C, Keenan D, Norgaard M. Evaluation of a pinna compensation algorithm for sound localization and speech perception in noise. Am J Audiol. 2013;22(6):84-93.
- Kuk F, Keenan D. Frequency transposition: Training is only half the story. Hearing Review. 2010;17(11):38-46.
- Sweetow R, Palmer C. Efficacy of individual auditory training in adults: a systematic review of the evidence. J Am Acad Audiol. 2005;16(7):494-504.
- Preece J. Sound localization by cochlear implant users. Semin Hear. 2010;31(1):37-46.
- Tyler RS, Witt SA, Dunn CC, et al. Initial development of a spatially separated speech- in-noise and localization training program. J Am Acad Audiol. 2010;21:390-403.
- Sweetow R, Henderson-Sabes J. The need for and development of an adaptive listening and communication enhancement (LACE) program. J Am Acad Audiol. 2006;17:538-558.
- Kumpik D, Kacelnik O, King A. Adaptive reweighting of auditory localization cues in response to chronic unilateral earplugging in humans. J Neuroscience. 2010;30(14):4883-4894.
- Amitay S, Halliday L, Taylor J, et al. Motivation and intelligence drive auditory perceptual learning. PLoS ONE. 2010;5(3):1-8.
- Irving S, Moore D. Training sound localization in normal hearing listeners with and without a unilateral ear plug. Hear Res. 2011;280(1-2):100-8.
- Edmondson-Jones A, Irving S, Moore D, et al. Planar localization analyses: A novel application of a center of mass approach. Hear Res. 2010;267:4-11.
- Wright B, Wilson R, Sabin A. Generalization lags behind learning on an auditory perceptual task. J Neurosci. 2010;30(35):11635-11639.
Original citation for this article: Kuk F, Crose B. Localization 102: Essentials of a Home-based Front-back Localization Training Program. Hearing Review. 2015;21(1):20-25.