Nearest neighbor search (NNS) among large-scale and high-dimensional vectors plays an important role in recent large-scale multimedia search applications. This paper proposes an optimized multiple codebook construction method for an approximate NNS scheme based on product quantization, where sets of residual sub-vectors are clustered according to their distribution and the codebooks for product quantization are constructed from these clusters. Our approach enables us to adaptively select the number of codebooks to be used by trading between the search accuracy and the amount of memory available.

Hybrid FEC/ARQ, which is a mixture of forward error correction (FEC) and automatic repeat request (ARQ), is a well-known technique aiming for packet-loss recovery to guarantee quality of service (QoS) for real-time communications. In this paper, focusing on layered video transmission over wireless network environment, we propose a proactive retransmission scheme for hybrid FEC/ARQ. In the proposed scheme, a receiver host periodically sends probe packets to a sender host in order to check wireless channel state. If the sender host does not receive any probe packet during a pre-specified interval, it regards the wireless channel as being in burst loss state, and it proactively retransmits packets expected to be lost during the burst loss period. The buffer management associated with layered video coding is also taken into consideration. The performance of the proposed scheme is investigated by simulation. Numerical examples show that the proposed scheme transmits packets of the base layer more successfully than the conventional FEC/ARQ.

Acoustic energy harvesters that function in environments where sound pressure is extremely high (150 dB), such as in engine rooms of aircraft, are expected to be capable of powering wireless health monitoring systems. This paper presents the power generation performance of a lead-zirconate-titanate (PZT) acoustic energy harvester with a vibrating PZT diaphragm. The diaphragm had a diameter of 2 mm, consisting of Al (0.1 µm)/PZT (1 µm)/Pt (0.1 µm)/Ti (0.1 µm)/SiO2 (1.5 µm). The harvester generated a power of 510-14 W under a sound pressure level of 110 dB at the first resonance frequency of 6.28 kHz. It was found that the generated power was increased to 2.010-13 W using a sound-collecting Helmholtz resonator cone with a height of 60 mm. The cone provided a Helmholtz resonance at 5.8 kHz, and the generated power increased from 9.710-15 W to 7.310-13 W at this frequency. The cone was also effective in increasing the bandwidth of the energy harvester.

Cross-phase modulation (XPM) induced by residual intensity modulation in coherent optical frequency-shift-keying (FSK) frequency division multiplexing (FDM) transmission systems that use dispersion-shifted fibers is evaluated theoretically and experimentally in terms of spectral profile deformation. The bit-error rate is measured in a 2.5-Gbit/s 4-channel 40-km dispersion-shifted fiber transmission experiment, and we confirm experimentally and theoretically that the power penalty in the presence of residual intensity modulation of over 4 mWp-p exceeds 1dB. Experimental results show that the penalty due to XPM is large even when the power of the newly generated lights caused by four-wave mixing is 20-dB less than that of signals. This confirms that residual intensity modulation must be reduced in continuous-phase (CP)-FSK-FDM systems even though they are designed to avoid generating four-wave mixing.

Bag of Visual Words (BoVW) is an effective framework for image retrieval. Query expansion (QE) further boosts retrieval performance by refining a query with relevant visual words found from the geometric consistency check between the query image and highly ranked retrieved images obtained from the first round of retrieval. Since QE checks the pairwise consistency between query and highly ranked images, its performance may deteriorate when there are slight degradations in the query image. We propose Query Bootstrapping as a variant of QE to circumvent this problem by using the consistency of highly ranked images instead of pairwise consistency. In so doing, we regard frequently co-occurring visual words in highly ranked images as relevant visual words. Frequent itemset mining (FIM) is used to find such visual words efficiently. However, the FIM-based approach requires sensitive parameters to be fine-tuned, namely, support (min/max-support) and the number of top ranked images (top-k). Here, we propose an adaptive support algorithm that adaptively determines both the minimum support and maximum support by referring to the first round's retrieval list. Selecting relevant images by using a geometric consistency check further boosts retrieval performance by reducing outlier images from a mining process. An important parameter for the LO-RANSAC algorithm that is used for the geometric consistency check, namely, inlier threshold, is automatically determined by our algorithm. We further introduce tf-fi-idf on top of tf-idf in order to take into account the frequency of inliers (fi) in the retrieved images. We evaluated the performance of QB in terms of mean average precision (mAP) on three benchmark datasets and found that it gave significant performance boosts of 5.37%, 9.65%, and 8.52% over that of state-of-the-art QE on Oxford 5k, Oxford 105k, and Paris 6k, respectively.