Towards Input Device Satisfaction Through Hand Anthropometry

Abstract: We collected the hand anthropometric data of 91 respondents to come up with a Filipino-based measurement to determine the suitability of an input device for a digital equipment, the standard PC keyboard. For correlation purposes, we also collected other relevant information like age, height, province of origin, and gender, among others. We computed the percentiles for each finger to classify various finger dimensions and identify length-specific anthropometric cut-points. We compared the percentiles of each finger dimension against the actual length of the longest key combinations when correct finger placement is used for typing, to determine whether the standard PC keyboard is fit for use by our sampled population. Our analysis shows that the members of the population with hand dimensions at extended position below 75th percentile and at 99th percentile are the ones who would most likely not reach the longest key combination for the left and the right hands, respectively. Using machine vision and image processing techniques, we automated the anthropometric process and compared the accuracy of its measurements to that of manual process'. We compared the measurement generated by our automated anthropometric process with the measurements using the manual one and we found out that they have a very minimal absolute difference. The data collected from this study could be used in other studies such as determining a good design for mobile and other handheld devices, or input devices other than keyboard. The automated method that we developed could be used to easily measure hand dimensions given a digital image of the hand and could be extended for measuring the entire human body for various other applications.

• The paper presents a machine vision-based automated anthropometry system that significantly reduces the manual measurement errors common in traditional surveys.
• The methodology integrates detailed image processing techniques with factorial ANOVA, achieving sub-centimeter discrepancies compared to manual measurements.
• The findings highlight the necessity for region-specific device designs, ensuring ergonomic suitability and enhanced accessibility for Filipino users.

Input Device Design Optimization via Automated Hand Anthropometry

Background and Motivation

This paper addresses the critical gap in anthropometric data guiding the design of input devices for Filipinos, with an emphasis on standard PC keyboards. It recognizes the influence of localized population-specific hand dimensions on ergonomic suitability, user satisfaction, and device accessibility. The paper identifies persistent accuracy and consistency shortcomings in manual anthropometric surveys—arising both from surveyor variability and time-dependent measurement drift—hence motivating the development of a machine vision-based, automated measurement framework.

Methodology

Anthropometric Survey Design

Ninety-one college students from diverse Filipino regions constitute the sample, yielding demographic stratification by urban/rural origin, gender, age, height, and weight. Manual measurements followed protocols analogous to Greiner (1992), capturing seven linear measures per hand: five finger lengths (thumb, index, middle, ring, pinky) and two pinky-index distances in relaxed and extended poses (PIR and PIE).

Manual Measurement Consistency Analysis

A two-factor factorial ANOVA evaluated measurement reliability over repeated sessions by multiple surveyors. Results indicate statistical significance ($\alpha = 0.001$) for both surveyor-dependent (inter-rater) and time-dependent (intra-rater) variability. Mean absolute discrepancy of at least 0.1 cm across surveyors was observed even under strict adherence protocols.

Automated Measurement System

Hand images were acquired with an embedded 5cm-square reference marker. Image processing pipeline comprised grayscale conversion, denoising, thresholding, binary morphological operations (erosion, dilation), and boundary extraction. Fingertips and valleys were algorithmically detected via local maxima analysis along the boundary distance curve from a wrist reference point. Finger lengths were computed as the distance from tip to baseline midpoint, normalized using the reference square for metric conversion. The pipeline achieves a mean absolute measurement error below 0.73 cm versus manual measurements.

Automated Anthropometry Pipeline (Pseudocode)

def automated_anthropometry(image, reference_square_coords): # Preprocessing: grayscale, median filter, thresholding img_gray = to_grayscale(image) img_denoised = median_filter(img_gray) img_binary = threshold(img_denoised) # Morphology: erosion, dilation img_cleaned = erode(img_binary) img_final = dilate(dilate(img_cleaned)) # Boundary extraction boundary_coords = extract_boundary(img_final) # Fingertip/valley detection wrist_ref = compute_wrist_reference(boundary_coords) dist_curve = calc_distances(boundary_coords, wrist_ref) extrema_points = detect_extrema(dist_curve, threshold=sigma) # Baseline estimation via valley/fingertip/isosceles triangle heuristic baselines = estimate_baselines(extrema_points) finger_lengths_px = [distance(tip, baseline_mid) for tip, baseline_mid in baselines] # Metric conversion using reference square scaling_factor = compute_scaling(reference_square_coords) finger_lengths_cm = [length * scaling_factor for length in finger_lengths_px] return finger_lengths_cm

Results

Demographic Data and Finger Length Percentiles

The measured finger lengths and PIR/PIE distances exhibit substantial variation across gender, urbanization, and age groups. The percentile analysis (1st, 5th, 25th, 50th, 75th, 95th, 99th) is critical for translating anthropometric data into design cut-points. Notably, individuals below the 75th percentile for left-extended hand and below 99th percentile for right-extended hand are unable to comfortably reach the widest two-key combinations (e.g., SHIFT-5, SHIFT-6) available on a standard 45 cm keyboard. This signals a misalignment between imported keyboard designs and the Filipino population's hand biomechanics.

Automated vs Manual Measurement Performance

Mean differences between automated and manual measures range from 0.37 cm to 0.73 cm across fingers and hands, with standard deviations indicating similar variance profiles. The system demonstrates operational suitability for large-scale surveys, substantially reducing human-induced measurement drift, fatigue artifacts, and throughput limitations.

Measurement Comparison Table (excerpt)

Finger	Mean (Actual)	Mean (Automated)	Mean Absolute Difference
Right Thumb (F1)	5.95 cm	5.48 cm	0.55 cm
Left Index (F2)	7.02 cm	7.43 cm	0.62 cm

Discussion and Implications

The findings underscore the necessity for regionally-adapted input device dimensions, invalidating one-size-fits-all paradigms often adopted by multinational hardware distributors. Ergonomic shortfalls for sub-75th percentile hand spans corroborate the practical exclusion of a significant user subset from optimal device interaction. The machine vision system offers scalable, cross-population anthropometric data acquisition, mitigating surveyor biases while vastly enhancing consistency.

This approach enables iterative hardware adaptation cycles based on anthropometric quantiles, supports rapid prototyping of custom device classes (keyboards, mobile devices), and fosters inclusive design for populations with unique hand morphology distributions.

Trade-Offs and Future Directions

While the automated measurement system attains sub-centimeter error rates, further algorithmic refinement (e.g., advanced fingertip detection, model-based hand pose estimation) may reduce errors for borderline percentile cases. The methodology can generalize to multi-joint limb measurements or full body anthropometry applications, supporting domains from device design to clinical diagnostics.

Scalability considerations include computational resource allocation for real-time or batch image processing, as well as user interface integration for mass anthropometric surveys.

Conclusion

This study advances the input device design discipline via robust, automated anthropometric measurement tailored to the Filipino context. By empirically demonstrating manual measurement unreliability and automating the process within sub-centimeter error margins, it provides a replicable framework for anthropometry-driven ergonomics. Population-informed percentile analysis reveals critical device accessibility deficits for current keyboard designs, highlighting the imperative for data-driven, inclusive interfaces.

Future research may focus on broader body part measurement automation and integration of large-scale anthropometric databases to inform device customization for diverse populations.