To collect infrared (IR) absorbance spectra from an automotive paint chip with an IR imaging microscope, it is a common practice to cast the paint chip in epoxy and then cross section it using a microtome to reveal the individual layers of paint. Ideally, the epoxy should present little or no spectral interference. However, the epoxy can infiltrate individual layers of the paint chip as it cures contaminating the IR spectra of the layers and impairing the accuracy of a search of each of these layers against an automotive paint library. In this study, we have demonstrated that automotive paint chips can be successfully cross sectioned without the use of embedding media. Sample preparation is easier, and more importantly, interfering peaks in the spectra due to the epoxy are eliminated. To demonstrate the advantages of this approach for sample preparation, IR image maps of four automotive paint chips that were not cast in epoxy prior to cross sectioning were collected. After each IR image was unfolded using an oblique transit to traverse the image, the spectra of the individual paint layers comprising the line map were reconstructed by alternating least squares. Comparing each recovered IR spectrum against a spectral library, we show that high quality spectral matches were obtained for spectra from the same line/model of the vehicle from which the paint sample originated. When the same paint chips were cast in epoxy prior to cross sectioning, high quality spectral matches could not always be obtained.
The application of Raman spectroscopy and pattern recognition methods to the problem of discriminating edible oils by type was investigated. Two-hundred and eighty-six Raman spectra obtained from 53 samples spanning 15 varieties of edible oils were collected for 90 s at 2 cm–1 resolution. Employing a Whittaker filter, all Raman spectra were baseline corrected after removing the high-intensity fluorescent background in each spectrum. The Raman spectral data were then examined using the three major types of pattern recognition methodology: mapping and display, discriminant development and clustering. The 15 varieties of edible oils could be partitioned into five distinct groups based on their degree of saturation and the ratio of polyunsaturated fatty acids to monounsaturated fatty acids. Edible oils assigned to one group could be readily differentiated from those assigned to other groups, whereas Raman spectra within the same group more closely resembled each other and therefore would be more difficult to classify by type.
Two‐hundred and fifteen Raman spectra of 15 edible oils or blends of edible oils from 53 samples spanning multiple brands purchased over 3 years were investigated using a genetic algorithm for spectral pattern recognition. Using a hierarchical approach to classification, the 15 edible oils could be divided into two groups based on their degree of unsaturation. While edible oils from any particular batch within a class are well clustered and can be differentiated from other varieties of edible oils that are also from a single source, incorporating uncontrolled variability from sources (by purchasing edible oils under different brand names) and seasons (by purchasing edible oils over a 3‐year period) presented a far more challenging classification problem for edible oils within the same group. The between‐source and yearly variability within one class of edible oils is often comparable to differences between the average spectra of the different varieties of edible oils, thereby preventing either a reliable classification of the edible oils or the detection of adulterants in an edible oil if a single model, spanning all sources and years of oils, is to be constructed. The novelty of this study arises from the incorporation of edible oils gathered systematically over the span of 3 years, introducing a heretofore unseen variance to the chemical compositions of the edible oils that are being classified. This is the first time that many different edible oils and commercially available brands thereof have been classified simultaneously.
Alternate least squares (ALS) reconstructions of the infrared (IR) spectra of the individual layers from original automotive paint were analyzed using machine learning methods to improve both the accuracy and speed of a forensic automotive paint examination. Twenty-six original equipment manufacturer (OEM) paints from vehicles sold in North America between 2000 and 2006 served as a test bed to validate the ALS procedure developed in a previous study for the spectral reconstruction of each layer from IR line maps of cross-sectioned OEM paint samples. An examination of the IR spectra from an in-house library (collected with a high-pressure transmission diamond cell) and the ALS reconstructed IR spectra of the same paint samples (obtained at ambient pressure using an IR transmission microscope equipped with a BaF2 cell) showed large peak shifts (approximately 10 cm−1) with some vibrational modes in many samples comprising the cohort. These peak shifts are attributed to differences in the residual polarization of the IR beam of the transmission IR microscope and the IR spectrometer used to collect the in-house IR spectral library. To solve the problem of frequency shifts encountered with some vibrational modes, IR spectra from the in-house spectral library and the IR microscope were transformed using a correction algorithm previously developed by our laboratory to simulate ATR spectra collected on an iS-50 FT-IR spectrometer. Applying this correction algorithm to both the ALS reconstructed spectra and in-house IR library spectra, the large peak shifts previously encountered with some vibrational modes were successfully mitigated. Using machine learning methods to identify the manufacturer and the assembly plant of the vehicle from which the OEM paint sample originated, each of the twenty-six cross-sectioned automotive paint samples was correctly classified as to the “make” and model of the vehicle and was also matched to the correct paint sample in the in-house IR spectral library.
Transformers have become popular in building end-to-end automatic speech recognition (ASR) systems. However, transformer ASR systems are usually trained to give output sequences in the left-to-right order, disregarding the right-to-left context. Currently, the existing transformer-based ASR systems that employ two decoders for bidirectional decoding are complex in terms of computation and optimization. The existing ASR transformer with a single decoder for bidirectional decoding requires extra methods (such as a self-mask) to resolve the problem of information leakage in the attention mechanism This paper explores different options for the development of a speech transformer that utilizes a single decoder equipped with bidirectional context embedding (BCE) for bidirectional decoding. The decoding direction, which is set up at the input level, enables the model to attend to different directional contexts without extra decoders and also alleviates any information leakage. The effectiveness of this method was verified with a bidirectional beam search method that generates bidirectional output sequences and determines the best hypothesis according to the output score. We achieved a word error rate (WER) of 7.65%/18.97% on the clean/other LibriSpeech test set, outperforming the left-to-right decoding style in our work by 3.17%/3.47%. The results are also close to, or better than, other state-of-the-art end-to-end models.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.