FastWavelet-Based Visual Classification

Abstract

We investigate a biologically motivated approach to fast visual classification, directly inspired by the recent work [13]. Specifically, trading-off biological accuracy for computational efficiency, we explore using standard wavelet transforms and patch transforms to parallel the tuning of visual cortex V1 and V4 cells, alternated with max operations to achieve scale and translation invariance. A feature selection procedure is applied during learning to accelerate recognition. We introduce a simple attention-like feedback mechanism, significantly improving recognition and robustness in multiple-object scenes. In experiments, the proposed algorithm achieves or exceeds state-of-the-art performance in object recognition, but also in new applications such as texture classification, satellite image classification, and language identification. Preliminary results on sound classification are shown as well.

Environmental sound recognition: A survey

“…This has inspired the development of visual features derived from the spectrogram of music signals [12], [43], [44]. The original application in [43] was texture classification, yet the plausible use for music instrument classification was mentioned. Souli and Lachiri subsequently used this method for ESR in [26].…”

Environmental sound classification using log-Gabor filter

“…This approach consists in computation of 12 log-Gabor filters from the environmental sounds spectrograms, with 2 different scales (1,2) and 6 different orientations (1,2,3,4,5,6), this extraction allows the best correlate of signal structures. Then, for each single filter result we calculated the magnitude, after that, we passed through on mutual information (MI) algorithm to find an optimal feature vector that can next be passed for classification phase [9].…”

“…Recently, little efforts emerge in the visual domain [4], [5].These studies, based on the visual signature extracted from the sound's spectrogram, and demonstrate that the time-frequency representation techniques can be applied to musical sounds, or environmental sounds and can produce a good result for classification. To improve our work results realized in [5], the idea consists in elaborating other methods based also on time-frequency representation, but with new approaches.…”

Audio classification from time-frequency texture

“…, M , The feature coefficient C[m] is expected to be discriminative if the time-frequency block B m contains some salient time-frequency structures. In this paper, we apply a simple random sampling strategy to learn the blocks as in [14,19]: each block is extracted at a random position from the logspectrogram S of a randomly selected training audio sample. Blocks of various sizes are applied to capture time-frequency structures at different orientations and scales [17].…”

“…Inspired by the recent biologically-motivated work on object recognition by Poggio, Serre and their colleagues [14], and more specifically on its variant [19] which has been shown to be particularly efficient for texture classification, we propose a simple feature extraction scheme based on timefrequency block matching (the effectiveness of application of time-frequency blocks in audio processing has been shown in previous work [17,18]). Despite its simplicity, the proposed algorithm relying only on visual texture features achieves surprisingly good performance in musical instrument classification experiments.…”