Results 1  10
of
6,750
MIME (Multipurpose Internet Mail Extensions) Part One: Mechanisms for Specifying and Describing the Format of Internet Message Bodies
 RFC 1521, BELLCORE, INNOSOFT
, 1993
"... RFC 822 defines a message representation protocol which specifies considerable detail about message headers, but which leaves the message content, or message body, as flat ASCII text. This document redefines the format of message bodies to allow multipart textual and nontextual message bodies to b ..."
Abstract

Cited by 403 (19 self)
 Add to MetaCart
RFC 822 defines a message representation protocol which specifies considerable detail about message headers, but which leaves the message content, or message body, as flat ASCII text. This document redefines the format of message bodies to allow multipart textual and nontextual message bodies
Measuring the information content of stock trades
 Journal of Finance
, 1991
"... This paper suggests that the interactions of security trades and quote revisions be modeled as a vector autoregressive system. Within this framework, a trade's information effect may be meaningfully measured as the ultimate price impact of the trade innovation. Estimates for a sample of NYSE is ..."
Abstract

Cited by 469 (11 self)
 Add to MetaCart
of the formal models. In the process, the framework establishes a rich characterization of the dynamics by which trades and quotes interact. The market considered here is a specialist market in which a designated marketmaker exposes bid and ask quotes to the trading public. An extensive theory has evolved
Momentum strategies
 Journal of Finance
, 1996
"... We examine whether the predictability of future returns from past returns is due to the market's underreaction to information, in particular to past earnings news. Past return and past earnings surprise each predict large drifts in future returns after controlling for the other. Market risk, si ..."
Abstract

Cited by 334 (4 self)
 Add to MetaCart
past performance. The results suggest a market that responds only gradually to new information. AN EXTENSIVE BODY OF RECENT finance literature documents that the crosssection of stock returns is predictable based on past returns. For example, DeBondt and Thaler (1985, 1987)report that longterm past
RHex: A simple and highly mobile hexapod robot
, 2001
"... In this paper, the authors describe the design and control of RHex, a power autonomous, untethered, compliantlegged hexapod robot. RHex has only six actuators—one motor located at each hip— achieving mechanical simplicity that promotes reliable and robust operation in realworld tasks. Empirically ..."
Abstract

Cited by 314 (53 self)
 Add to MetaCart
stable and highly maneuverable locomotion arises from a very simple clockdriven, openloop tripod gait. The legs rotate full circle, thereby preventing the common problem of toe stubbing in the protraction (swing) phase. An extensive suite of experimental results documents the robot’s sig
Simple BM25 Extension to Multiple Weighted Fields
, 2004
"... This paper describes a simple way of adapting the BM25 ranking formula to deal with structured documents. In the past it has been common to compute scores for the individual fields (e.g. title and body) independently and then combine these scores (typically linearly) to arrive at a final score for t ..."
Abstract

Cited by 213 (11 self)
 Add to MetaCart
This paper describes a simple way of adapting the BM25 ranking formula to deal with structured documents. In the past it has been common to compute scores for the individual fields (e.g. title and body) independently and then combine these scores (typically linearly) to arrive at a final score
ParticleBased Fluid Simulation for Interactive Applications
, 2003
"... Realistically animated fluids can add substantial realism to interactive applications such as virtual surgery simulators or computer games. In this paper we propose an interactive method based on Smoothed Particle Hydrodynamics (SPH) to simulate fluids with free surfaces. The method is an extension ..."
Abstract

Cited by 280 (11 self)
 Add to MetaCart
Realistically animated fluids can add substantial realism to interactive applications such as virtual surgery simulators or computer games. In this paper we propose an interactive method based on Smoothed Particle Hydrodynamics (SPH) to simulate fluids with free surfaces. The method is an extension
Experimental evaluation of wireless simulation assumptions
, 2004
"... All analytical and simulation research on ad hoc wireless networks must necessarily model radio propagation using simplifying assumptions. A growing body of research, however, indicates that the behavior of the protocol stack may depend significantly on these underlying assumptions. The standard res ..."
Abstract

Cited by 260 (12 self)
 Add to MetaCart
All analytical and simulation research on ad hoc wireless networks must necessarily model radio propagation using simplifying assumptions. A growing body of research, however, indicates that the behavior of the protocol stack may depend significantly on these underlying assumptions. The standard
Narmour, 1990; Schoenberg, 1967; Toch, 1948/1977) analyses of musical structure and has given rise to extensive bodies of research on melody identification
"... In two experiments, descriptions of melodic contour structure and predictions of perceived similarity relations between pairs of contours produced by a number of different models are examined. Two of these models, based on the musictheoretic approaches of Although numerous psychologists and music ..."
Abstract
 Add to MetaCart
In two experiments, descriptions of melodic contour structure and predictions of perceived similarity relations between pairs of contours produced by a number of different models are examined. Two of these models, based on the musictheoretic approaches of Although numerous psychologists and music theorists have evinced an interest in contour, only a few formal descriptions of contour structure have been proposed. One approach to studying contour has been to categorize melodies according to feature information (Adams, 1976; Kolinski, 1965; Seeger, 1960; Polansky & Bassein, 1992). For example, Adams (1976) describes a "melodic contour typology" in which contours are defined by a set of boundary properties (the initial, final, highest, and lowest pitches within the contour), with three possible relations existing between boundary pitches (greater than, equal to, and less than). Combinations of these boundaries and relations are then used to identify primary contour features such as overall melodic slope, and secondary contour features such as repetitions of the highest or lowest pitch (see In a similar vein, Morris (1993) describes a model of contour relations in which a given contour is transformed through the reiterative application of a reducing algorithm into its fundamental or prime form. This algorithm repeatedly deletes notes from the contour, ultimately focusing in on the initial, final, maximum, and minimum points of the contour. According to Morris, these primes "play a role analogous to that of the Schenkerian background" (p. 218) and thus function as the underlying contour of the phrase in question. Similarity relations between prime forms of different contours can then be determined and used to highlight structural equivalence relations between different contours. Although this approach moves beyond the more generic taxonomic category schemes of other approaches (e.g., The individual pitch relations within the CC define a series of contour intervals (CIs) that describe the relative pitch intervals between adjacent and nonadjacent notes within the CC. For example, the CI for the adjacent interval 1 » 2 of the first melody in Numerous possible quantifications can be derived by using this information; two such measures are the mean pitch interval size and the summed pitch interval size for all intervals in the contour, with these values charac3. It would also be possible to count intervals in contour space, using the same contour coding procedure as that of Friedmann and Marvin and Laprade. In fact, these two methods give highly comparable results. 302 Mark A. Schmuckler terizing oscillation somewhat differently. Because mean pitch interval information is normalized by the number of contour reversals, it treats each ascending and descending motion within the contour equally; thus, mean pitch interval is related to individual interval size within the contour. In contrast, because summed pitch interval information aggregates across all intervals, it provides a global measure of pitch divergence within the contour. For both interval measures (as well as for the reversals), similarity is calculated using difference scores for these values. The distinction between the two pitch interval measures can be appreciated by considering the mean and summed interval information for the two sample contours. The first sample contour has a mean pitch interval value of 26.5 semitones, whereas the second contour has a value of 12.5 semitones. In contrast, the first sample contour has a summed interval value of 53 semitones, whereas the second contour has a value of 49 semitones. Thus, the summed pitch interval measure suggests a greater similarity between the contours than does the mean pitch interval measure. An alternative characterization of the global shape of a contour involves looking for the presence of repeated or cyclical patterns. For example, the contours of One typical procedure for quantifying cyclical information involves the application of Fourier analysis techniques. Described simply, Fourier analysis converts a signal from the temporal domain into the frequency domain, providing a mathematical decomposition of the signal into a set of harmonically related sine waves. Each of these sine waves is characterized by an amplitude (strength) and phase (timing relation) value, with the relative amplitudes and phases of the sine wave components providing a quantitative measure of the different cyclical patterns within the signal. Considering all these frequency components, Fourier analysis provides a general description of the shape of the contour, taking into account both slow moving, lowfrequency movement such as general trend, as well as any high frequency, pointtopoint fluctuation in the contour. Fourier analysis has a number advantages for contour analysis. One strength is that it provides an obvious measure of similarity contours with comparable cyclical structure are perceived as related by listeners. The degree of association between contours can be assessed in any number of ways. One procedure involves correlating amplitude and phase spectra. A correlation coefficient is attractive in that it compares only relative patterns of ups and downs, collapsing across absolute differences in the spectra, and has associated tables of statistical significance. It is possible, however, that the magnitude of differences in the spectra are meaningful. To capture this aspect, one can calculate an absolute difference score between corresponding components of the spectra, with contour similarity based on the amount of difference between individual frequency components, or on an average difference calculated across all harmonics. Regardless of the 304 Mark A. Schmuckler associative measure used, the implication is that the degree of association between contours will be related to the perception of contour similarity. A second advantage is that the output of a Fourier analysis is largely invariant in the face of different transformations of the contour. For example, Fourier analysis is essentially scale independent, such that a contour consisting of large interval leaps will produce comparable phase and amplitude spectra to a similar contour made up of small steps. In the same vein, the lowfrequency components of the output of a Fourier analysis will be relatively invariant in the face of melodic ornamentation, such as grace notes, passing tones, and neighbor tones; higher frequency components will, however, vary somewhat. In both of these cases, the coding of the input representation becomes important. For scale independence, different codings produce outputs that can vary from highly similar to identical. As for melodic ornamentation, the degree of similarity between outputs will vary depending on whether or not the input representation contains rhythmic information, although both rhythmic and nonrhythmic inputs should produce highly related outputs; further consideration of rhythm in contour analysis is taken up in the general discussion. Irrespective of these final issues, it seems clear that Fourier analysis of contour is a potentially powerful technique. Applying Fourier procedures to the analysis of melodic contours is not without some analytic pitfalls, however. One serious concern is that Fourier analysis requires a variety of assumptions, some of which are violated in this application. For example, one important assumption is that the series being analyzed via Fourier techniques is an extended sequence that represents the sampling of an infinitely periodic signal; in contrast, melodic contours are short and discrete. This concern has both theoretical and practical implications. Theoretically, the concern is that because the series to be analyzed does not meet this criterion, the results of the analysis, in terms of its ability to forecast future cyclical patterns, is suspect. Practically, the concern is that with a short, discrete series, a Fourier analysis will be disproportionately influenced by information at the beginning and ending of the sequence, with such "edge effects" distorting the Fourier spectra, also resulting in decreased predictive power. In defense of the theoretical viability of this application, it should be remembered that the goal here is not the typical predictive forecasting commonly associated with timeseries analyses (the family of analytic procedures to which Fourier analysis belongs). In contrast, Fourier analysis is being used strictly as a convenient tool to quantify the presence of repetitive patterns in a melody. Fundamentally, Fourier analysis is simply a mathematical decomposition procedure that is applicable to any numerical series. Although care must be taken in interpreting the output of this procedure, Testing Models of Melodic Contour Similarity 305 there is no reason why this technique cannot be used to provide a description of any numerical series, including a melodic contour. As for the practical considerations, one procedure used to control edge effects involves mathematically transforming the series by using windowing techniques (e.g., Hamming or Kaiser windows). Such windows reduce edge effects by weighting the beginning and ending of a sequence less than the middle of the sequence. Unfortunately, given the length of the series tested in this work, such procedures actually alter the shape of the contours themselves; thus, similarity measures based on such altered contours are of questionable utility. As such, it is ultimately difficult to control edge effects, meaning that the results of a Fourier analysis of contour may be too distorted to characterize the melody adequately. The goal of these studies was to test the ability of these models to predict perceived contour similarity. Toward this end, all of these models quantified individual contours and predicted similarity relations between pairs of contours. These similarity measures were then compared with a listenergenerated derived measure of perceived contour similarity. The decision to use a derived measure of similarity, as opposed to gathering direct similar306 Mark A. Schmuckler ity ratings of pairs of melodies, was in large part pragmatic. Pilot work suggested that listeners had great difficulty providing direct similarity ratings for pairs of melodies (using the stimulus melodies of Experiment 1, described later), with listeners frequently reporting that they had forgotten the first melody by the end of the second melody. In keeping with these reports, there was little intrasubject or intersubject reliability for contour similarity ratings. Thus, an alternative procedure for generating similarity data was used, involving gathering contour ratings of individual melodies, and then intercorrelating aggregate ratings to produce a derived similarity matrix. This approach is a widely accepted alternative to a direct similarity rating procedure Procedure Listeners were told they were participating in an experiment on contour perception. They were informed that on each trial they would hear a 12note contour and that they should rate the complexity of this contour, using a 9point scale, with 1 indicating not very complex and 9 indicating very complex. Listeners typed their responses on the computer keyboard, and as soon as their response was entered, the computer began the next trial. Listeners were free to take breaks between the blocks of trials. Subsequent to the final block of trials, listeners completed a musical background questionnaire and were debriefed as to the purpose of the experiment. The entire experiment lasted approximately 3045 min. RESULTS The first step in the analysis involved quantifying contour similarity on the basis of the various models. All of these models produce multiple measures of similarity. Friedmann (1985) provides differences in the CASV, the CCVI, and the CCVII; because they are difference scores, these values represent dissimilarity, with small numbers reflecting similarity and large numbers dissimilarity. Marvin and Laprade (1987) measure similarity via CSIM and CMEMB values. For both of these theories, measures are calculated between two contours in prime form, as well as between one contour in prime form and the second contour in retrograde, inversion, or retrogradeinversion form. Thus, contour similarity can be based on similarity between any of these forms, or the maximum similarity value across these forms; in fact, Marvin and Laprade (1987, pp. 237, 245) assume that contour similarity is a function of this "maximum" value. Fourier analysis also produces multiple similarity measures, including correlations of the amplitude or phase spectra for pairs of contours, as well as difference scores between the harmonic components for the amplitude and phase spectra. Finally, contour oscillation also produces multiple similarity values, involving difference scores based on the number of contour reversals, the mean pitch interval, and the summed pitch interval. Again, because these are differences they produce dissimilarity values, and hence should be negatively related to perceived similarity. sures.4 4. The CMEMB values were actually calculated as the total percent overlap, aggregating across CSUBSEGs of length 2 to 11. It is also, however, possible to calculate this value as an average percent overlap for CSUBSEGs of length 2 to 1 1 individually. In fact, all analyses for the CMEMB model were calculated in both ways and provided virtually identical patterns. Accordingly, only the results for the total percent overlap are presented. For CASV, CCVI, and CCVII models, these values represent the average absolute difference score between corresponding (e.g., positive and negative) digits in the vectors. Finally, regression techniques fit the theoretical models to contour similarity;
Is there universal recognition of emotion from facial expression? A review of the crosscultural studies
 Psychological Bulletin
, 1994
"... Emotions are universally recognized from facial expressions—or so it has been claimed. To support that claim, research has been carried out in various modern cultures and in cultures relatively isolated from Western influence. A review of the methods used in that research raises questions of its eco ..."
Abstract

Cited by 213 (4 self)
 Add to MetaCart
or soft hair, but many share his beliefs about facial expression and emotion. Oatley and Jenkins (1992) observed, "By far the most extensive body of data in the field of human emotions is that on facial expressions of emotion" (p. 67). Recent reviews of those data (see Table 1) agree
Progressive search space reduction for human pose estimation
 In CVPR
, 2008
"... The objective of this paper is to estimate 2D human pose as a spatial configuration of body parts in TV and movie video shots. Such video material is uncontrolled and extremely challenging. We propose an approach that progressively reduces the search space for body parts, to greatly improve the chan ..."
Abstract

Cited by 226 (30 self)
 Add to MetaCart
shot, by softlabeling every pixel as belonging to a particular body part or to the background. We demonstrate upperbody pose estimation by an extensive evaluation over 70000 frames from four episodes of the TV series Buffy the vampire slayer, and present an application to fullbody action recognition
Results 1  10
of
6,750