Enrollment Session Research Articles

Abstract B088: The Alabama Hereditary Cancer Cohort – New strategies for African American recruitment

Abstract Our group has established the Alabama Hereditary Cancer Cohort and primarily investigates breast cancer genetics in African American, an underrepresented group in biomedical research. Alabama has double the national percentage of African Americans and is the location of the infamous and unethical Tuskegee Syphilis Study. Therefore, strategic protocols were developed to break research participation barriers, effectively recruiting African American hereditary breast cancer cases and families. Our initial recruitment strategies involved both hospital-based and community-based recruitment; the latter involved our recruitment bus, the Gene Machine, and focused on African American engagement. Since our initial publication, our community-based recruitment protocol has been modified. Many of the modifications were due to the pandemic, limiting in-person community engagement. In-person education seminars and enrollment sessions transitioned to virtual meetings, using various video-conferencing platforms. Regarding virtual enrollments, upon agreement to proceed with the virtual consent process, study participants are mailed an enrollment kit containing study consent forms, saliva collection kits, return mailing materials, and instructions. Once received, an enrollment appointment is made, and the consent process follows the procedure for in-person study enrollment, concluding with the saliva sample collection. Once collected, the sample is packaged with the consent forms and returned to our laboratory in the postage-paid mailer. Community-based recruitment has also been enhanced by partnering with the Alabama Department of Public Health to identify study participants through the Alabama Statewide Cancer Registry, similar to the very successful Carolina Breast Cancer Study. Furthermore, Research Champions have been added to the protocol, defined as primary healthcare providers and specialists who aid in recruitment by identifying individuals who fit the study criteria. The concept of Research Champions was developed after recognizing that the requirements to execute hospital-based recruitment are arduous on partnering hospitals. Thus, the sole function of a Research Champion is identifying prospective study participants, relieving the burden on physicians and hospital staff. Lastly, our Gene Machine Facebook page and newly designed website, tailored towards laypeople/study participants, facilitate additional engagement through ads, videos, and photos. Ultimately, these processes reduce the travel burden, aid in social distancing, and expand our reach, enhancing recruitment and enrollment efforts and breaking barriers to African American research. Citation Format: Nancy Merner, Elizabeth Stallworth, Erica Reasor, Katia Khoury, Lily Gutnik, Meagan Farmer, Brandon Johnson. The Alabama Hereditary Cancer Cohort – New strategies for African American recruitment [abstract]. In: Proceedings of the 15th AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2022 Sep 16-19; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2022;31(1 Suppl):Abstract nr B088.

DeepEyedentificationLive: Oculomotoric Biometric Identification and Presentation-Attack Detection Using Deep Neural Networks

We study involuntary micro-movements of both eyes, in addition to saccadic macro-movements, as biometric characteristic. We develop a deep convolutional neural network that processes binocular eye-tracking signals and verifies the viewer’s identity. In order to detect presentation attacks, we develop a model in which the movements are a response to a controlled stimulus. The model detects replay attacks by processing both the controlled but randomized stimulus and the ocular response to this stimulus. We acquire eye movement data from 150 participants, with 4 sessions per participant and conduct experiments on this new and legacy data sets with varying tracker precision and sampling rate. We observe that the model detects replay attacks reliably. For identification and identity verification, the model attains substantially lower error rates than prior work. We explore the relationships between training population size, training data volume, types of visual stimuli, number of training and enrollment sessions, interval between enrollment and probe sessions on one hand and the model performance on the other hand.

Robot-mediated overground gait training for transfemoral amputees with a powered bilateral hip orthosis: a pilot study

BackgroundTransfemoral amputation is a serious intervention that alters the locomotion pattern, leading to secondary disorders and reduced quality of life. The outcomes of current gait rehabilitation for TFAs seem to be highly dependent on factors such as the duration and intensity of the treatment and the age or etiology of the patient. Although the use of robotic assistance for prosthetic gait rehabilitation has been limited, robotic technologies have demonstrated positive rehabilitative effects for other mobility disorders and may thus offer a promising solution for the restoration of healthy gait in TFAs. This study therefore explored the feasibility of using a bilateral powered hip orthosis (APO) to train the gait of community-ambulating TFAs and the effects on their walking abilities.MethodsSeven participants (46–71 years old with different mobility levels) were included in the study and assigned to one of two groups (namely Symmetry and Speed groups) according to their prosthesis type, mobility level, and prior experience with the exoskeleton. Each participant engaged in a maximum of 12 sessions, divided into one Enrollment session, one Tuning session, two Assessment sessions (conducted before and after the training program), and eight Training sessions, each consisting of 20 minutes of robotically assisted overground walking combined with additional tasks. The two groups were assisted by different torque-phase profiles, aiming at improving symmetry for the Symmetry group and at maximizing the net power transferred by the APO for the Speed group. During the Assessment sessions, participants performed two 6-min walking tests (6mWTs), one with (Exo) and one without (NoExo) the exoskeleton, at either maximal (Symmetry group) or self-selected (Speed group) speed. Spatio-temporal gait parameters were recorded by commercial measurement equipment as well as by the APO sensors, and metabolic efficiency was estimated via the Cost of Transport (CoT). Additionally, kinetic and kinematic data were recorded before and after treatment in the NoExo condition.ResultsThe one-month training protocol was found to be a feasible strategy to train TFAs, as all participants smoothly completed the clinical protocol with no relevant mechanical failures of the APO. The walking performance of participants improved after the training. During the 6mWT in NoExo, participants in the Symmetry and Speed groups respectively walked 17.4% and 11.7% farther and increased walking speed by 13.7% and 17.9%, with improved temporal and spatial symmetry for the former group and decreased energetic expenditure for the latter. Gait analysis showed that ankle power, step width, and hip kinematics were modified towards healthy reference levels in both groups. In the Exo condition metabolic efficiency was reduced by 3% for the Symmetry group and more than 20% for the Speed group.ConclusionsThis study presents the first pilot study to apply a wearable robotic orthosis (APO) to assist TFAs in an overground gait rehabilitation program. The proposed APO-assisted training program was demonstrated as a feasible strategy to train TFAs in a rehabilitation setting. Subjects improved their walking abilities, although further studies are required to evaluate the effectiveness of the APO compared to other gait interventions. Future protocols will include a lighter version of the APO along with optimized assistive strategies.

The effect of remote patient monitoring on the primary care clinic visit frequency among adults with type 2 diabetes

AimsHealthcare organizations are increasingly using technology to assist in diabetes management based on telemedicine’s proven ability to improve glycemic regulation, decrease cost, and overcome barriers to effective healthcare. Nevertheless, it remains unclear how telemedicine intersects with primary care. We aim to measure the impact of a remote monitoring program for diabetes on primary care delivery through analysis of primary care office visit frequency. MethodsPatients eligible to participate in our institution’s remote diabetes monitoring program were identified and classified as enrolled or not enrolled (i.e. “usual care”). The number of scheduled and completed primary care office visits in the 12 months prior to and after the index date were measured for both groups. The index date was the enrollment date or, for the patients who received usual care, the next available enrollment session after eligibility screen. Two-sample t-tests were used to examine the change in frequency of office visits prior to and after enrollment for participants, as well as the difference in visit frequency between enrolled patients versus patients receiving usual care. ResultsThere was no statistical difference in the number of scheduled or completed primary care clinic visits before or after enrollment in telehealth. Furthermore, there was no difference in the number of scheduled or completed primary care visits between patients enrolled in telehealth versus those receiving usual care. ConclusionParticipation in telehealth has been shown to be associated with significant HbA1c reductions in prior work, yet our data suggest that remote monitoring is not associated with a change in primary care office visit frequency. This suggests that telehealth may improve diabetes management independently of primary care visits.

Self-Reported Quality of Life Outcomes in Aphasia Using Life Participation Approach Values: 1-Year Outcomes

ObjectiveTo evaluate the effectiveness of a newly initiated aphasia center program using Life Participation Approach to Aphasia (LPAA) values. DesignEvidence-based practice quality improvement project. SettingAmbulatory care aphasia center part of a larger health care system. ParticipantsParticipants included 41 clients with aphasia and 40 primary caregivers of clients with aphasia. Participants included all clients and caregivers that were enrolled in the aphasia center within the first year of opening. All participants were enrolled in 1-4 eleven-week sessions of the aphasia center program. Participants with aphasia included 19 women and 22 men with a mean age of 65.2 years (range, 33-84 years). InterventionsGroup therapy was provided using the LPAA statement of values with each client receiving 3 hours of treatment once per week for 11 weeks. Clients were provided with the opportunity to complete multiple 11-week enrollment sessions if they chose to do so. Main Outcome MeasuresClient-reported outcomes included the Stroke and Aphasia Quality of Life Scale, and caregiver-reported outcomes included the Modified Caregiver Strain Index. Both measures are self-reported and scale based. Client satisfaction data were also collected via a self-reported program evaluation upon completion of each 11-week session. ResultsResults indicated statistically significant improvements in quality of life scores for clients with aphasia following initial participation of an 11-week aphasia center program (P≤.0001) and following 1-year participation (P≤.0001). Additionally, caregiver’s level of burden decreased significantly following initial participation of the family member (P=.003) and following completion of three 11-week sessions (P≤.0001). Self-reported client satisfaction program evaluation results also demonstrated improved quality of life (100%) and improved language skills (average=89.75%). ConclusionsThe project provided further evidence supporting the use of LPAA values with clients experiencing chronic aphasia.

Parental interest in genomic sequencing of newborns: enrollment experience from the BabySeq Project

PurposeNewborn genomic sequencing (nGS) has great potential to improve pediatric care. Parental interest and concerns about genomics are relatively unexplored. Understanding why parents decline research consent for nGS may reveal implementation barriers. MethodsWe evaluated parental interest in a randomized trial of nGS in well-baby and intensive care unit nursery settings. Interested families attended an informational enrollment session (ES) with a genetic counselor prior to consenting. Reason(s) for declining participation and sociodemographic associations were analyzed. ResultsOf 3860 eligible approached families, 10% attended ES, 67% of whom enrolled. Of 1760 families queried for decline reasons, 58% were uninterested in research. Among 499 families considering research, principal reasons for decline prior to ES included burdensome study logistics (48%), feeling overwhelmed postpartum (17%), and lack of interest/discomfort with genetic testing (17%). Decliners after ES more often cited concerns about privacy/insurability (41%) and uncertain/unfavorable results (23%). ConclusionLow interest in research and study logistics were major initial barriers to postpartum enrollment and are likely generic to many postpartum research efforts. Concerns over privacy and result implications were most commonly cited in decliners after ES. Understanding parental concerns around research nGS may inform future integration of nGS into newborn screening, predictive testing, and pediatric diagnostics.

Biometric touchstroke authentication by fuzzy proximity of touch locations

Advancing touchscreen technologies lead to deployment of biometric passwords that could provide an additional security layer, therefore the presence of a secondary and hidden ghost password could restrain incoming fraud attacks even if the main password is stolen. Apart from the traditional keystroke methods dealing with the inter-key times, on-screen keyboards potentially have new features to extract. Provided that the touchscreen devices capture the touch locations, the users could intentionally create and save a ghost password by touching the predefined regions of the keys in enrollment session. However, while logging in, it is not easy to touch exactly the same points compared to the coordinates saved in enrollment; touching closer coordinates in each attempt is still possible though. From this viewpoint, we introduce a fuzzy classifier with the mathematical foundations of fuzzy proximity and inference for a novel location-based authentication system running on an emulated on-screen keyboard. We mainly focused on extracting the global coordinates that the user is touching in the two-step enrollment, which is so common in conventional registration interfaces. As the main classifier, a fuzzy inference system is implemented with proximity control of the touches for registration and login sessions. The results of the classification procedure are greatly encouraging that only one of sixty four fraud attacks was inadvertently granted; while the false reject rate is 0% and the false accept rate is 1.56% with the equal error rate is 1.61% which indeed represents one of the lowest among the classifiers introduced before

Automatic animation of an articulatory tongue model from ultrasound images of the vocal tract

Visual biofeedback is the process of gaining awareness of physiological functions through the display of visual information. As speech is concerned, visual biofeedback usually consists in showing a speaker his/her own articulatory movements, which has proven useful in applications such as speech therapy or second language learning. This article presents a novel method for automatically animating an articulatory tongue model from ultrasound images. Integrating this model into a virtual talking head enables to overcome the limitations of displaying raw ultrasound images, and provides a more complete and user-friendly feedback by showing not only the tongue, but also the palate, teeth, pharynx, etc. Altogether, these cues are expected to lead to an easier understanding of the tongue movements. Our approach is based on a probabilistic model which converts raw ultrasound images of the vocal tract into control parameters of the articulatory tongue model. We investigated several mapping techniques such as the Gaussian Mixture Regression (GMR), and in particular the Cascaded Gaussian Mixture Regression (C-GMR) techniques, recently proposed in the context of acoustic-articulatory inversion. Both techniques are evaluated on a multispeaker database. The C-GMR consists in the adaptation of a GMR reference model, trained with a large dataset of multimodal articulatory data from a reference speaker, to a new source speaker using a small set of adaptation data recorded during a preliminary enrollment session (system calibration). By using prior information from the reference model, the C-GMR approach is able (i) to maintain good mapping performance while minimizing the amount of adaptation data (and thus limiting the duration of the enrollment session), and (ii) to generalize to articulatory configurations not seen during enrollment better than the GMR approach. As a result, the C-GMR appears to be a good mapping technique for a practical system of visual biofeedback.

Sentence‐HMM state‐based i‐vector/PLDA modelling for improved performance in text dependent single utterance speaker verification

In this paper, we make use of hidden Markov model (HMM) state alignment information in i-vector/probabilistic linear discriminant analysis (PLDA) framework to improve the verification performance in a text-dependent single utterance (TDSU) task. In the TDSU task, speakers repeat a fixed utterance in both enrollment and authentication sessions. Despite Gaussian mixture models (GMMs) have been the dominant modeling technique for text-independent applications, an HMM based method might be better suited for the TDSU task since it captures the co-articulation information better. Recently, powerful channel compensation techniques such as joint factor analysis (JFA), i-vectors and PLDA have been proposed for GMM based text-independent speaker verification. In this study, we train a separate i-vector/PLDA model for each sentence HMM state in order to utilize the alignment information of the HMM states in a TDSU task. The proposed method is tested using a multi-channel speaker verification database. In the experiments, it is observed that HMM state based i-vector/PLDA (i-vector/PLDA-HMM) provides approximately 67% relative reduction in equal error rate (EER) when compared to the i-vector/PLDA. The proposed method also outperforms the baseline GMM and sentence HMM methods. It yields approximately 51% relative reduction in EER over the best performing sentence HMM method.

Using Gene Expression Analysis to Examine Changes in Loneliness, Depression and Systemic Inflammation in Lonely Chronically Ill Older Adults.

PurposeThe purpose of this study was to evaluate the effectiveness of LISTEN (Loneliness Intervention) on loneliness, depression, physical health, systemic inflammation, and genomic expression in a sample of lonely, chronically ill, older adults.MethodsThis was a prospective, longitudinal randomized trial of LISTEN, a novel intervention based on theories of narrative and cognitive restructuring to target specific aspects of loneliness. Twenty-three older, lonely, chronically ill adults were recruited from a family medicine clinic in West Virginia. Participants were randomized to two groups, 13 in LISTEN group (Loneliness Intervention) and 10 in attention control (healthy aging education). Participants attended an enrollment session where they completed consent, survey data (including sociodemographics and chronic illness diagnoses), baseline physical measures, and blood sampling for gene expression analysis. After completing the 5 weekly sessions, all participants attended a 12 week post data collection meeting (17 weeks post-baseline) for survey completion, physical measures and blood sampling.ResultsThe results of this study show that the LISTEN intervention improves measures of physical and psychosocial health. Specifically, subjects enrolled in LISTEN showed reductions in systolic blood pressure, as well as decreased feelings of loneliness and depression. These changes may be due, in part, to a reduction in systemic inflammation, as measured by interleukin-2.ConclusionThis study provides support for the use of LISTEN in reducing loneliness in chronically ill, older adults. Further, while some of our results are inconclusive, it provides rationale to expand our study population to evaluate the relationship between loneliness and systemic inflammation. In the future, enhancing knowledge about the relationships among loneliness, chronic illness, systemic inflammation, and gene expression of these particular targets, and how these relationships may change over time with intervention will inform translation of findings to clinical settings.

WAVE Pilot Study: Feasibility of Using Emails and Short Message Service (SMS) with High School Soccer Players to Reinforce Compliance in an Obesity Prevention Intervention

ObjectiveEvaluate the feasibility of using emails/SMS with high school (HS) soccer players for increasing compliance at group classes, completion of questionnaires, and 7‐day tracking of eating behavior and physical activity. 
 
MethodCustomized emails/SMS using Google Voice and WavePipe (a custom data collection system that sends emails) were sent to remind and encourage compliance in 52 youth (14‐19y) enrolled in the WAVE Pilot Study. Prior to each activity, coaches and youth received emails/SMS about enrollment session, class schedule (intervention group [IG] only), and baseline assessments. During a subsequent 7‐day tracking period, youth received a daily email/SMS with a unique link to the behavioral checklist (BC) online and a reminder to wear the Fitbit.ResultsOverall, the IG was more compliant in all aspects of data collection and attending group sessions. The enrollment and survey completion rate was 76% in the IG (n=25) vs. 51% in the control group (CG) (n=27) (p<0.05). On average, the IG wore Fitbits for 6.0 d vs. 4.9 d for the CG (p<0.05); 98% Fitbits were returned. The IG completed the online BC for 3.7 d vs. 1.4 d by the CG (p<0.05). Failure to receive daily emails significantly reduced the BC response rate; 44% of the CG reported not noticing 蠅1 daily email/SMS. Other reasons include wrong contacts. Youth generally preferred SMS over emails as reminders. 
ConclusionEmails, particularly SMS, during intervention increased active HS soccer players' compliance. 
 
 Sponsor: NIFA AFRI Award No. 2013‐67001‐20418

Evaluation of a Worksite Wellness Program Designed to Reduce Cardiovascular Risks

We evaluated a multifactorial worksite wellness program designed to improve lifestyle and reduce cardiovascular risks at a small professional services company. Program participation (N = 60 employees) consisted of an enrollment session, a 6-month period of wellness program activities, and a disenrollment session. Lifestyle and biometric measures were obtained at enrollment and disenrollment. Over 6-months, percentage of dietary fat intake decreased (P < .01); daily fruit and vegetable servings and fiber intake increased (P = .02); systolic and diastolic blood pressure decreased (P = .04 and 0.01, respectively), and high-density lipoprotein increased (P < .01). This worksite wellness program achieved meaningful improvements in dietary intake and reduction in cardiovascular risks.

PS2-17: Diabetes Social Support Feasibility Pilot Study: Utilizing Mobile Technology and Self-Identified Supporters to Enhance Self-Monitoring of Blood Glucose

Background and Aims: Self-monitoring of blood glucose (SMBG) is associated with improved glycemic control among patients with type 2 diabetes, however, the practice of daily self-monitoring is not optimal. Telecommunications technology may improve adherence to recommended self-management practices by remotely transmitting automated reminders to motivate patients, and utilizing social networking for peer support. The purpose of this pilot study is to demonstrate the feasibility and usability of mobile technology and the potential added value of social support to improve SMBG frequency and glycemic control among adults with type 2 diabetes. Methods: Adults 25–74 years of age with type 2 DM and an average HbA1c > 8.0% were recruited from Kaiser Permanente Georgia (KPGA) and Oakhurst Medical Center (OMC, a community health clinic) to participate in a 3-month study using wireless technology. Enrollment sessions with presentations on SMBG techniques, use of the wireless technology, and motivational coaching to enhance social support were conducted in November 2009. During the subsequent 3-months, both diabetes patients and their self-selected supporters will receive text messages to their cell phones summarizing a patient’s SMBG frequency and levels. Participants and their supporters will attend a disenrollment session in February 2010 when feasibility and usability will be assessed in focus groups. Results: 6 of 161 eligible diabetes patients at KPGA and 9 of 28 eligible diabetes patients at OMC, and their self-selected supporters, consented to participate. The average age of diabetes patients was 49.3 years. 86.7% (N=13) were African-American; and 33.3% (N=5) were male. Five days after enrollment, 60% (N=9) of patients had connected their wireless transmitters and had current blood glucose data. Follow-up phone calls will be made to ensure that all participants are connected to the wireless technology within 10 days of the enrollment session. Conclusion: Integrating mobile telecommunications technology with chronic disease management may empower patients in their own self-care and ease the burden on health care providers. Our study will evaluate the potential for studying the use of wireless mobile technology in a larger randomized controlled trial and will obtain participant comments on what changes might improve participant compliance.

C-A1-02: Evaluating Kaiser Permanente Georgia's Worksite Wellness Programs in the Atlanta Area

Background/Aims: There are few prospective evaluations of the effectiveness of worksite wellness programs in achieving improved lifestyle and health. Our study has two objectives: to assess the cross-sectional associations of biometric status, lifestyle, and health obtained at a wellness program’s enrollment session, and to assess changes in biometric status, lifestyle, and health between the enrollment and 6-month disenrollment sessions. Methods: For this prospective cohort study, participants were recruited from Kaiser Permanente Georgia (KPG) subscribers who were employed at a mid-sized private employer group. Participants completed a biometric assessment and written survey. The biometric measures included height, weight, blood pressure, hip/waist ratio, and total cholesterol. The written survey included measures of: health (SF-12), patient activation (PAM-13), leisure physical activity (BRFSS), food screeners (Block), and work presenteeism (SPS-6) and absenteeism. Over the next 6 months, participants will attend seminars on exercise, weight loss, and healthy cooking and eating. The biometric assessment and survey be repeated in February 2010. Results: Of the 63 eligible KPG subscribers, 60 (41 females and 19 males) consented to participate. Participant ages ranged from 28–77 years. Overall, physical and emotional health, activation, leisure physical activity, dietary intake, and BMI were comparable to results obtained among relatively healthy, low risk working age adults in a 2005 survey of KPG subscribers in large group employers. Total cholesterol was significantly (p<0.05), positively associated with percent fat in diet; and, percent fat in diet was significantly, inversely associated with activation. Participants with 1 or more days absent from work in the most recent 4-week period reported significantly higher total cholesterol and percent fat in diet and significantly lower activation. Participants with high levels of presenteeism were significantly less likely to report leisure physical activity at the recommended level. Conclusions: The initial results indicate that improving levels of activation (improved self-efficacy, motivation to pursue a healthy lifestyle and adhere to recommended care) might improve dietary intake (thereby reducing fat intake and cholesterol levels) and decrease BMI. The final wellness program evaluation will indicate how much change in these measures might occur in a relatively healthy group of working age adults.

Predictors of Enrollment in a Smoking Cessation Clinical Trial After Eligibility Screening

A fraction of smokers considered eligible for pharmacotherapy treatment trials for nicotine dependence enrolled. Little is known about smokers not enrolling in a treatment trial or how these smokers differ from those who have enrolled. We screened 2257 individuals for a smoking cessation trial involving behavioral counseling and a novel medication. Of those screened, 33% of callers were eligible for enrollment (N=753). Of those eligible for the trial, 37% attended the subsequent enrollment session (N=282). We compared the 282 attendees to the 471 smokers who were eligible for the trial but did not attend. Logistic regression indicated that African American smokers were about half as likely to enroll in the trial (odds ratio [OR], 0.55; 95% CI, 0.36-0.84); for every year increase in age, participants were 4% more likely to enroll (OR, 1.04; 95% CI, 1.02-1.06); and participants who were motivated to enroll in the trial for financial incentives were 42% less likely to enroll in the trial (OR, 0.58; 95% CI, 0.45-0.76). These findings suggest the need to devise and test recruitment strategies directed toward African Americans and younger smokers to increase enrollment among eligible smokers to smoking cessation treatment trials.

System for voice verification of telephone transactions

A system and a method is disclosed for verifying a voice of a user conducting a telephone transaction. The system and method includes a mechanism for prompting the user to speak in a limited vocabulary. A feature extractor converts the limited vocabulary into a plurality of speech frames. A pre-processor is coupled to the feature extractor for processing the plurality of speech frames to produce a plurality of processed frames. The processing includes frame selection, which eliminates each of the plurality of speech frames having an absence of words. A Viterbi decoder is also coupled to said feature extractor for assigning a frame label to each of the plurality of speech frames to produce a plurality of frame labels. The processed frames and frame labels are then combined to produce a voice model, which includes each of the plurality of frame labels that correspond to the number of plurality of processed frames. A mechanism is also provided for comparing the voice model with the claimant's voice model, derived during a previous enrollment session. The voice model also is compared with an alternate voice model set, derived during previous enrollment sessions. The identity claimed is accepted if the voice model matches the claimant's voice model better than the alternative voice model set.

Speaker verification using randomized phrase prompting

Speaker verification is a viable method of controlling access to computer and communications networks and physical places. The system described here is capable of accurately verifying an individual’s claimed identity from a short sample of his or her speech. Design requirements for the system included the following: (1) an enrollment session should take no longer than 3 min; (2) an acceptance or rejection decision should be reached within 20 s; (3) the probabilities of false rejection and false acceptance should be as low as possible; and (4) prompts should be chosen at random from a large number of possibilities. A number of systems have been designed to satisfy the first three requirements [l-3], but do not adequately address requirement (4). Several systems use a single fixed word or phrase as a “voice password” [4,5]. To address requirement (4), the system described here uses a prompting strategy in which phrases are composed at random using a small vocabulary of words. The spoken phrases are compared with word templates derived from enrollment sessions. This strategy introduces the following difficulty. Words occur in the test material in contexts that did not occur in the enrollment material. The context in which a word is spoken influences its pronunciation through coarticulation, caused by limitations in the movement of the speech articulators. These unmodeled coarticulations contribute to the measured dissimilarity between the input speech and the claimant’s word templates, increasing the likelihood of rejecting valid users. A scoring method called likelihood ratio scoring partially overcomes this difficulty. Likelihood ratio scoring is based on an approximation to the likelihood ratio, the ratio of the probability of the observed input assuming it was spoken by the claimant to the probability of the observed input assuming it was spoken by someone else. To compute this approximation, input speech is compared with the claimant’s templates and with the templates of a set of other speakers (who may be other enrolled users) called the ratio set. All templates are derived from enrollment sessions in which the same scripts were read; therefore, the same contexts are represented in the templates. When a new context is encountered in test material, all raw match scores to templates are affected similarly. The likelihood ratio score, however, by virtue of its definition as a ratio, is relatively unaffected. The template matching algorithm, which is otherwise conventional [6], is modified for computation of the likelihood ratio score. In particular, a specialized syntax is used, and additional data from the template match, other than the “recognized” word sequence, are saved. The syntax is constructed to make certain that the algorithm is quite robust with respect to false starts and nonspeech sounds. A reliable enrollment method, in which speaker-specific word templates are computed automatically using several incremental steps, starting with speaker-pooled templates for male and female speakers, was developed. Two real-time prototype systems were developed using Sun-3 workstations with additional processor boards. One such system was tested, simultaneously collecting a large speaker verification database. A rationale was developed for determining the size of the test required to allow hypotheses regarding the system’s true error rates to be tested with stated confidence levels. The remainder of the paper is organized as follows. The speech material used for verification is described in Section 2. The algorithm is then presented in Sec-