Citation Context

...ver, to our understanding, these models have no interpretation in information theory. Our metric shares with Fitts-TP the Gaussian channel interpretation of movement as a limited transmission channel =-=[4]-=-: “information capacity is limited only by the amount of statistical variability, or noise, that is characteristic of repeated efforts to produce the same response” [6: p. 262]. In our metric, changes...

Citation Context

...erview of computation steps in calculating information capacity (TP) in full-body movement. APPROACH The concept of information capacity here follows the Gaussian channel interpretation of Paul Fitts =-=[6,7,12]-=- but applies it to mutual information I(x; y) in movement sequence x and its repetition y. I(x; y) denotes the reduction in bits in entropy of x when y is known. Since our I(x; y) excludes most of the...

Citation Context

...iation of the mutual information estimate obtained by plugging the empirical correlation coefficient into Eq. (10) can be characterized by means of results from classical likelihood ratio test theory =-=[3,8]-=-. In the case ρ = 0, the estimator asymptotically follows a χ 2 distribution scaled by a factor 0.5 or, equivalently, a Gamma distribution Γ (k = 1 / 2, θ = 1), and in the case ρ ≠ 0, its distribution...

Citation Context

...ble. Information theory has contributed to the measurement of user performance in HCI by providing a metric that collapses data on speed and accuracy into a single metric: throughput (bits/s, or bps) =-=[7,12,13,21]-=-. Throughput is often measured as statistical variability in aimed movements wherein the user brings an end-effector (pointer) on top of a spatially expanded target. Information capacity denotes the r...

Citation Context

... We have learned that, most of the time, dimension reduction with PCA preserves the order of TPs and can be used if absolute throughput values are unimportant. For computation, we utilize Fast GP-LVM =-=[10]-=- to transform the two sequences and obtain two new time series, r (x′) and r (y′) . Each frame in the new sequences represents a latent variable corresponding to a frame in the original sequence. Figu...

Citation Context

...enotes the rate at which the user could have sent messages, given her speed and accuracy for given target properties. Selecting targets with the mouse, for instance, yields throughputs of 3.7–4.9 bps =-=[17]-=-. Although the metric has been contested, no better alternatives exist for comparing performance across tasks, conditions, and devices. This paper extends the measurement of throughput from aimed move...

Citation Context

...preferred solution is Gaussian Process Latent Variable Modeling (GP-LVM)[16] administered separately on the residuals of x and y. GP-LVM models have been used to model human movement, such as walking =-=[18]-=-. In our experience, GP-LVM provides more effective dimension reduction than does Principal Component Analysis (PCA), which is limited to linear relationships. Using GP-LVM typically reduced TPs by a ...

Citation Context

...n the movie Minority Report (Figure 11). The case is relevant to HCI, because such interfaces are touted without regard for the fact that bimanual continuous control suffers from interference effects =-=[14]-=-. As the critical condition we study the hand position’s effect. Inspired by the movie, we assume a user gesturing with both hands raised to a space of 120° of the field of view. With this manipulatio...

Citation Context

...ble. Information theory has contributed to the measurement of user performance in HCI by providing a metric that collapses data on speed and accuracy into a single metric: throughput (bits/s, or bps) =-=[7,12,13,21]-=-. Throughput is often measured as statistical variability in aimed movements wherein the user brings an end-effector (pointer) on top of a spatially expanded target. Information capacity denotes the r...

Citation Context

...cording to observed inaccuracies: The effective width is defined via the distribution of offsets from target center We = 4.133 σ [13,17]. Extensions of Fitts’ law models to continuous aimed movements =-=[1]-=- covered only path width and length originally but were later extended to curvature [11]. However, to our understanding, these models have no interpretation in information theory. Our metric shares wi...

Citation Context

...er interfaces that effectively map human movement to virtual movement. To assess joint human–computer performance, the “tempting but naïve” solution is to examine average speed and accuracy in a task =-=[21]-=-. This approach, however, overlooks the fact that data from easy and from difficult motor acts are incommensurable. Information theory has contributed to the measurement of user performance in HCI by ...

Citation Context

...ents in Step II, but because swaying is poorly matched from one sequence to another, even with CTW, TPs are negligible. This is not to say that balancing would not be motorically difficult, for it is =-=[5]-=-. Fitness games such as Nintendo Wii Fit Balance Board measure variation during balancing for score calculation. However, for an external observer such as an interface, it carries no information after...

Citation Context

...MSE Robustness to Noise in Motion Capture Most studies of aimed movements have employed measurement instruments in which the level of noise is low. Noise is inevitable in motion capture data, however =-=[20]-=-. Ideally, the metric would tolerate a level of noise that does not obscure controlled aspects of movement. To understand the effect of noise, we added white noise with zero mean and variance to each ...

Citation Context

...iation of the mutual information estimate obtained by plugging the empirical correlation coefficient into Eq. (10) can be characterized by means of results from classical likelihood ratio test theory =-=[3,8]-=-. In the case ρ = 0, the estimator asymptotically follows a χ 2 distribution scaled by a factor 0.5 or, equivalently, a Gamma distribution Γ (k = 1 / 2, θ = 1), and in the case ρ ≠ 0, its distribution...

Citation Context

...orally align frames in x with frames in y. Our solution is to align sequence pairs with Canonical Time Warping (CTW), a state-of- the-art technique for aligning sequences that describe human behavior =-=[22]-=-. CTW uses the more traditional Dynamic Time Warping [1] as an initial solution but improves it by adopting features from Canonical Correlation Analysis. The result is a pairwise alignment of x and y,...

Citation Context

...on and Throughput Under models (5) and (6), the differential entropy of each sequence can be estimated by plugging the residual variance into the familiar formula for the Gaussian entropy: h(x) ≈ , , =-=(9)-=- where is the residual variance of x. n 2 log2(2πeσ ˆ 2 x ) h(y) ≈ n 2 log2(2πeσ ˆ 2 y ) n 2 (x) 2 ˆ σ x = (rt ) /n ∑ t =1 € The mutual information, € which yields the reduction in bits in the entropy...

Citation Context

... any complex movement that a user could produce with practice. The idea in such manipulations is that learned motor programs retain some invariance when transferred from a familiar context to another =-=[15]-=-. For example, one can sign one’s name with the teeth or behind one’s back. The effects of constraints such as position, rotation, or scale on TP show how robust the user’s movements are to the condit...

Citation Context

...igning interfaces with full-body control is that the number of possible movements is too enormous to study empirically. One alternative would be to sample the space of possible movements aggressively =-=[19]-=- and average TPs. Another is to impose constraints in order to expose human factors relevant to interface design. Our solution is to divide the interaction space into movement conditions and ask a use...

Citation Context

...f offsets from target center We = 4.133 σ [13,17]. Extensions of Fitts’ law models to continuous aimed movements [1] covered only path width and length originally but were later extended to curvature =-=[11]-=-. However, to our understanding, these models have no interpretation in information theory. Our metric shares with Fitts-TP the Gaussian channel interpretation of movement as a limited transmission ch...