Abstract: An SV reduction method for multi-class SVMs is provided with which a number of SVs contained in the multi-class SVMs can be reduced without becoming trapped in a local minimum optimization solution and the reduction of the SVs can be performed at high precision and high speed. The method includes a step of selecting, from a plurality of initially present support vectors, support vector pairs zi, zj (i, j=1, 2, . . . , NS); a step of preparing a single-variable objective function with a single global maximum and determining a maximum value k of the objective function; and a step of applying the maximum value k to the support vector pairs zi and zj to determine a temporary vector Ztemp[i] of small classification errors; and the support vector pairs zi, zj are represented by the temporary vector Ztemp[i].

Abstract: The present invention provides a terminal apparatus that charges an appropriate amount of money according to the usage of individual users without using a network. A terminal apparatus reproduces content in response to a reproduction request from a user. The terminal apparatus includes a charging management unit that charges for the reproduction of the content and stores charging history as history information; a history information database that stores the history information in association with the reproduced content; and a removal determining unit that determines whether to remove the reproduced content from a charging target on the basis of the magnitude relationship between the standard price of the content charged for the reproduction and the sum of charges based on the history information. In this way, it is possible to charge an appropriate amount of money according to the usage of individual users and achieve a user-oriented charging process.

Abstract: An audio playback apparatus includes a playback section for playing audio back, a buffer section (13) for temporarily storing voice data to be played back, a switch (14) for on/off switching transmission of the voice data from the buffer section (13) to the playback section, at least one threshold judgment section (18) for judging whether an amount of voice data buffered in the buffer section (13) is equal to or smaller than at least one threshold value, the at least one threshold judgment section (18) providing a true output when the buffered voice data amount is equal to or smaller than the at least one threshold value, a timer (19) triggered by an output from the at least one threshold judgment section to output a trigger pulse with a predetermined duration of time, and a first control section (191) for controlling on/off of the switch (14) to prevent transmission of the voice data in the buffer section (13) to the playback section for the predetermined duration of time of the trigger pulse when the true o

Abstract: An object of the present invention is to perform DCT coefficient operation and drift compensation for embedding watermark data with the small amount of computing. An MB data extraction unit extracts a macro block or block into which the watermark data is embedded on the basis of a template. Additionally, a DCT coefficient extraction unit extracts the DCT coefficient at a position into which the watermark data is embedded according to a direction of the template. A DCT coefficient operation unit performs the process of embedding the watermark data or the process of compensating the drift to the extracted DCT coefficient.

Abstract: It is made possible to concretely estimate actual transmission bit rate in a frame of grasping a line state by using a report packet. Together with media packets encoded by a video encoder 1 and an audio encoder 2, a sender report packet generated by a sender report generation section 3 is sent to the transmission line. A receiver report packet is fed back from the receiver side. Each of the sender report packet and the receiver report packet includes report packets of two kinds differing in size. A rate estimation section 4 estimates transmission bit rate on the basis of round-trip delay time for the sender report packet and the receiver report packet each having a small size and round-trip delay time for the sender report packet and the receiver report packet each having a large size.

Abstract: A re-learning method includes: a step of learning an SVM by using a set of training samples for initial learning which have known labels; a step of perturbation-processing the training samples for initial learning; a step of using the perturbation-processed sample as a training sample for addition; and a step of re-learning the learned SVM by using the training sample for initial learning and the training sample for addition. For the training samples for initial learning to be perturbation-processed, a training sample obtained by removing a training sample for initial learning corresponding to a non-support vector, a training sample corresponding to a support vector existing on a soft margin hyperplane, etc., may be used.

Abstract: A plural number of training vectors are randomly selected from a total of unused training vectors, and from among the selected training vectors, a vector having the largest error amount is extracted. Subsequently, the extracted vector is added to the already used training vector so as to update the training vector, and the updated training vector is used to learn the SVM. When the largest error amount becomes smaller than a certain setting value ? or when the already used training vector becomes larger than a certain value m, learning of a first phase is stopped. In learning of a second phase, the learning is performed on a predetermined number of or all of the training vectors having a large error amount.

Abstract: To provide a method capable of reducing support vectors without decreasing the performance of an SVM. The method includes: a step of learning an SVM by using a set of training samples for initial learning which have known labels; a step of evaluating a training sample for initial learning corresponding to an outlier (value greater than 0 and equal to or less than C) based on a parameter a value obtained by learning the SVM; and a step of removing the training sample for initial learning corresponding to the outlier from a set of the original training samples for initial learning.

Abstract: Re-encoding information indicating whether or not an input moving picture has been compression-encoded is provided to a mode determining unit. The mode determining unit determines an optimal encoding mode for the input moving picture, out of a plurality of encoding modes. The mode determining operation is controlled by the re-encoding information. According to a mode determination result of the mode determining unit, a prediction error value corresponding to the determined mode is selected by a selection switch, and then, an encoding process is performed. As the re-encoding information, a predetermined value, a bit rate, and a quantization parameter, etc., could be provided.

Abstract: An encoding unit encodes data at a first rate during an initial predetermined section of uncompressed data (a), and encodes data at a second rate after the initial predetermined section (where the first rate<the second rate), and encoded data are stored in a storage. The storage is read out the stored data to a network at a transmission rate equal to the transmission rate of the network. A decoding unit decodes received data at the first rate during an initial predetermined section, and decodes received data at a second rate after the initial predetermined section. For finite contents, receiving completion and decoding completion of the received data are simultaneous. In accordance with the present invention, while tolerance with respect to data incoming fluctuations is ensured, playback of dynamic images and music and the like can be immediately started after receiving.

Abstract: A video analyzing unit 1 analyzes features of an input video. A video classifying unit 2 estimates, based on results of an analysis by the video analyzing unit 1, whether the input video is one shot by a professional cameraman or one shot by an amateur to carry out classification. The video analyzing unit 1 includes a shot density measuring unit 11, a camera-shake determining unit 12, a blur determining unit 13, and a contrast measuring unit 14. Further, a sound analyzing unit that analyzes features of a sound accompanying the input video can be provided to use results of an analysis thereof as information for video classification.

Abstract: Degraded digital data is distributed from a server 1 to each user A, B, C, and D. An encrypted degradation compensation module is also delivered to the digital data duplicating apparatus 2 for the user A. The digital data duplicating apparatus 2 decrypts the encrypted degradation compensation module using its own encryption key, when playing digital data, to generate a 100% degradation compensation module. This module is applied to the degraded digital data to compensate the degradation thereof. An encrypted degradation compensation module of which the degradation compensation capability is reduced is delivered from the user A to the user B. The user B can use the encrypted degradation compensation module of which the degradation compensation capability is reduced from the user A to replicate the degraded digital data.

Abstract: A barcode-embedding area is obtained in a color image based on expression possibility by the juxtaposed additive color mixture and possibility of reading of pixels separately by a barcode reader (S11), a color image in the barcode-embedding area is expressed by the juxtaposed additive color mixture, and color images for the code area and the background area, which are readable by the barcode reader, are created (S12, S13), a color-image-superimposed barcode image is created by using the color images for the code area and the background area, and the same color-image-superimposed barcode image is embedded in the barcode-embedding area (S14), and the color-image-superimposed barcode image and the original color image are corrected if necessary (S15, S16).

Abstract: A second tally image generating means 20 that outputs a pixel value B1m,n of a second tally image B1 includes a decoding-time peripheral luminance value calculating section composed of a decoded pixel value calculating means 28 and a peripheral luminance value calculating means 29. The decoding-time peripheral luminance value calculating means 29 calculates a luminance value dm,n around a currently processed pixel. A confidential-image embedding means 26 calculates a pixel value B1m,n of a second tally image B1 by use of a temporarily determined pixel value B1tm,n, a luminance value Sm,n of a confidential image S, a pixel value B0m,n of a first tally image B0 and the peripheral luminance value dm,n.

Abstract: The present invention provides an electronic watermark editor for a compressed video that can efficiently carry out editing such as deletion or rewriting of electronic watermark data. DCT coefficient values are extracted while applying all templates to video contents, and a template where a total sum of relative absolute value differences between pairs of the DCT coefficient values is maximized is judged to be an applied template. The pairs of DCT coefficient values extracted by the applied template are processed by a smoothing process section so that the relative differences are reduced and so as to reduce individual relative differences recursively until significant differences between the coefficient value pairs are statistically eliminated. In addition, the pairs of DCT coefficient values extracted by the applied template are replaced with new editing bits.

Abstract: Separating means 11 separates a picture signal into a reference signal and one or more signals to be predicted. A reference signal including a quantization error is outputted from inverse converting means 18. Predicting means 14 estimates one or more prediction coefficients for approximating the signal to be predicted based on the reference signal including the quantization error. Compensating means 19 generates a prediction signal of the signal to be predicted. Difference calculating means 13 generates a prediction error signal. The prediction error signal is processed and outputted by the converting means 12, quantizing means 15 and encoding means 16.

Abstract: The present invention provides a scrambled compressed data distribution device that can perform a high-resistance scrambled compressed data distribution by selectively using a plurality of types of scramble information and a playback device thereof. A scrambling section scrambles compressed data by unique one of a plurality of scrambling methods. A hash value calculating section determines a hash value based on data b in a position where digital watermark data is not embedded, and an encrypting section encrypts the hash value with a secret key Sk(n) corresponding to the scrambling method and outputs an encrypted hash value c. A watermark data embedding section embeds the encrypted hash value in a watermark data embedding position of the data b and transmits the same as a compressed data stream d.

Abstract: First, a set of seed images that tally with each other are prepared (S101). Next, a set of threshold matrices are created from the set of seed images (S102) Next, a plurality of tally images are created from an input gray-scale image by using the set of threshold matrices (S103). The tally images are combined, whereby the confidential image is decoded.

Abstract: A transmitting device includes a first buffer for temporarily storing audio data, a second buffer for temporarily storing video data, a multiplexing unit for outputting audio data coming from the first buffer and video data coming from the second buffer onto a single signal channel, and a transmission buffer monitoring unit for monitoring the amount of first buffered data stored in the first buffer and, if the amount of the first buffered data exceeds a first threshold, instructing the second buffer to discard video data.

Abstract: An audio playback apparatus includes a playback section for playing audio back, a buffer section (13) for temporarily storing voice data to be played back, a switch (14) for on/off switching transmission of the voice data from the buffer section (13) to the playback section, at least one threshold judgment section (18) for judging whether an amount of voice data buffered in the buffer section (13) is equal to or smaller than at least one threshold value, the at least one threshold judgment section (18) providing a true output when the buffered voice data amount is equal to or smaller than the at least one threshold value, a timer (19) triggered by an output from the at least one threshold judgment section to output a trigger pulse with a predetermined duration of time, and a first control section (191) for controlling on/off of the switch (14) to prevent transmission of the voice data in the buffer section (13) to the playback section for the predetermined duration of time of the trigger pulse when the true o