Predictive Coding Algorithm Research Articles

Hierarchical predictive coding in distributed pain circuits.

Predictive coding is a computational theory on describing how the brain perceives and acts, which has been widely adopted in sensory processing and motor control. Nociceptive and pain processing involves a large and distributed network of circuits. However, it is still unknown whether this distributed network is completely decentralized or requires networkwide coordination. Multiple lines of evidence from human and animal studies have suggested that the cingulate cortex and insula cortex (cingulate-insula network) are two major hubs in mediating information from sensory afferents and spinothalamic inputs, whereas subregions of cingulate and insula cortices have distinct projections and functional roles. In this mini-review, we propose an updated hierarchical predictive coding framework for pain perception and discuss its related computational, algorithmic, and implementation issues. We suggest active inference as a generalized predictive coding algorithm, and hierarchically organized traveling waves of independent neural oscillations as a plausible brain mechanism to integrate bottom-up and top-down information across distributed pain circuits.

Automatic Speech to Sign Language Software for Translation (ASSiST) for Hearing-Impaired

Abstract: A total of 63 million people in India have hearing impairment, which is a common cause of disability. Due to communication barriers, these individuals are at risk for reduced cognitive skills and language deficits which may contribute to their poor general and oral health. Instructing these individuals regarding oral health concerns, treatment options and prognosis is often challenging for dental professionals. An Automatic Speech to Sign Language Translation Software (ASSiST) was developed to improve brushing technique among children with hearing impairment. The programming of Automatic Speech Recognition (ASR) system was done using Linear Predictive coding (LPC) algorithm for extraction of Speech Signals and Artificial Neural Networks as a classifier and recognizer. This tool is tested to help imparting oral hygiene instructions in the form of a sequence-based brushing technique to the children chair side which when spoken by the dentist translates it into the respective regional sign language which is displayed on the screen.

Reverse Differentiation via Predictive Coding.

Deep learning has redefined AI thanks to the rise of artificial neural networks, which are inspired by neuronal networks in the brain. Through the years, these interactions between AI and neuroscience have brought immense benefits to both fields, allowing neural networks to be used in a plethora of applications. Neural networks use an efficient implementation of reverse differentiation, called backpropagation (BP). This algorithm, however, is often criticized for its biological implausibility (e.g., lack of local update rules for the parameters). Therefore, biologically plausible learning methods that rely on predictive coding (PC), a framework for describing information processing in the brain, are increasingly studied. Recent works prove that these methods can approximate BP up to a certain margin on multilayer perceptrons (MLPs), and asymptotically on any other complex model, and that zerodivergence inference learning (Z-IL), a variant of PC, is able to exactly implement BP on MLPs. However, the recent literature shows also that there is no biologically plausible method yet that can exactly replicate the weight update of BP on complex models. To fill this gap, in this paper, we generalize (PC and) Z-IL by directly defining it on computational graphs, and show that it can perform exact reverse differentiation. What results is the first PC (and so biologically plausible) algorithm that is equivalent to BP in the way of updating parameters on any neural network, providing a bridge between the interdisciplinary research of neuroscience and deep learning. Furthermore, the above results in particular also immediately provide a novel local and parallel implementation of BP.

A Multitone Model-Based Seismic Data Compression

This work develops a model-based compression scheme for seismic data. First, seismic traces are modeled as multitone sinusoidal waves superposition. Each sinusoidal wave is regarded as a model component and is represented by a set of distinct parameters. Second, a parameter estimation algorithm for this model is proposed accordingly. In this algorithm, the parameters are estimated for each component sequentially. A suitable number of model components is determined by the level of the residuals energy. Next, the residuals are compressed using entropy coding or quantization coding techniques. The corresponding compression ratios are presented. Finally, the proposed model-based compression scheme is compared with the linear predictive coding (LPC) algorithm and the distributed principal component analysis (DPCA) algorithm on a real seismic database. The performance of the proposed model based is shown to be superior to that of the LPC and DPCA.

Multimodal Representation Learning for Place Recognition Using Deep Hebbian Predictive Coding.

Recognising familiar places is a competence required in many engineering applications that interact with the real world such as robot navigation. Combining information from different sensory sources promotes robustness and accuracy of place recognition. However, mismatch in data registration, dimensionality, and timing between modalities remain challenging problems in multisensory place recognition. Spurious data generated by sensor drop-out in multisensory environments is particularly problematic and often resolved through adhoc and brittle solutions. An effective approach to these problems is demonstrated by animals as they gracefully move through the world. Therefore, we take a neuro-ethological approach by adopting self-supervised representation learning based on a neuroscientific model of visual cortex known as predictive coding. We demonstrate how this parsimonious network algorithm which is trained using a local learning rule can be extended to combine visual and tactile sensory cues from a biomimetic robot as it naturally explores a visually aliased environment. The place recognition performance obtained using joint latent representations generated by the network is significantly better than contemporary representation learning techniques. Further, we see evidence of improved robustness at place recognition in face of unimodal sensor drop-out. The proposed multimodal deep predictive coding algorithm presented is also linearly extensible to accommodate more than two sensory modalities, thereby providing an intriguing example of the value of neuro-biologically plausible representation learning for multimodal navigation.

Analysing Predictive Coding Algorithms for Document Review

Abstract: Lawsuits and regulatory investigations in today's legal environment demand corporations to engage in increasingly intense data-focused engagements to find, acquire, and evaluate vast amounts of data. In recent years, technology-assisted review (TAR) has become a more crucial part of the document review process in legal discovery. Attorneys now have been using machine learning techniques like text classification to identify responsive information. In the legal domain, text classification is referred to as predictive coding or technology assisted review (TAR). Predictive coding is used to increase the number of relevant documents identified, while reducing human labelling efforts and manual review of documents. Deep learning models mixed with word embeddings have demonstrated to be more effective in predictive coding in recent years. Deep learning models, on the other hand, have a lot of variables, making it difficult and time-consuming for legal professionals to choose the right settings. In this paper, we will look at a few predictive coding algorithms and discuss which one is the most efficient among them. Keywords: Technology-assisted-review, predictive coding, machine learning, text classification, deep learning, CNN , Unscented Kalman Filter, Logistic Regression, SVM

A strategy for correcting errors in automated formant extraction

Sociophonetic vowel analysis relies heavily on measurements of resonant frequencies, particularly of the first and second formants. Automated formant estimation using linear predictive coding (LPC) algorithms in software like Praat greatly increases efficiency compared to hand measurements and allows researchers to analyze more data than was possible before this technological advancement. However, LPC analysis is prone to certain types of errors (e.g., Di Paolo et al., 2011; Harrison, 2013; Labov et al., 2006; Strelluf, 2019; Styler, 2017). In one common error, which I call “faulty low F2” (FLF2), LPC identifies a spectral peak between F1 and F2 as an F2 measurement. In automated extraction, the real F2 is then recorded as F3. Manually correcting these errors is time-consuming, potentially erasing the gains made by automated extraction. However, a systematic error suggests the possibility of a systematic solution. This presentation describes a method for identifying and correcting FLF2 errors using a script in R. The script identifies possible errors based on expected ranges and trajectories, determines whether the recorded F3 at the measurement point meets the criteria for a legitimate F2, and makes substitutions where appropriate. Results of tests with real data are presented. [Work supported by NSF #XXXXXXX.]

Predictive Coding: Adopting and Adapting Artificial Intelligence (AI) In Civil Litigation

This paper examines how predictive coding, an artificial intelligence (AI) technology, can effectively and efficiently complement the work of lawyers in the area of electronic discovery document review in civil litigation. It begins with a general overview of AI, and how machine learning can be used to automate the document review process in civil litigation. It then proceeds to a comprehensive overview of predictive coding technology and a discussion of legal issues related to the use of predictive coding technology in civil litigation. The legal issues are whether the use of artificial intelligence technology (as opposed to human intelligence) in document review complies with the rules of the court relating to documentary disclosure; and whether litigation privilege applies to seed sets (or training sets) used in training the predictive coding algorithm. Adopting a comparative law methodology, the paper seeks to address these issues. The paper concludes with a brief consideration of legal professionalism issues arising from the adoption of predictive coding technology in civil litigation in the context of Rule 3.1 of the Model Code of Professional Conduct dealing with competency. The paper argues that successful adoption of AI technology in civil litigation will extend the lawyer’s duty of competence to include knowledge of the relevant legal technology.

How the peak glottal area affects linear predictive coding-based formant estimates of vowels.

The estimation of formant frequencies from acoustic speech signals is mostly based on Linear Predictive Coding (LPC) algorithms. Since LPC is based on the source-filter model of speech production, the formant frequencies obtained are often implicitly regarded as those for an infinite glottal impedance, i.e., a closed glottis. However, previous studies have indicated that LPC-based formant estimates of vowels generated with a realistically varying glottal area may substantially differ from the resonances of the vocal tract with a closed glottis. In the present study, the deviation between closed-glottis resonances and LPC-estimated formants during phonation with different peak glottal areas has been systematically examined both using physical vocal tract models excited with a self-oscillating rubber model of the vocal folds, and by computer simulations of interacting source and filter models. Ten vocal tract resonators representing different vowels have been analyzed. The results showed that F1 increased with the peak area of the time-varying glottis, while F2 and F3 were not systematically affected. The effect of the peak glottal area on F1 was strongest for close-mid to close vowels, and more moderate for mid to open vowels.

Human Micro-Doppler Frequency Estimation Approach for Doppler Radar

Micro-Doppler frequency estimation is an important issue for radar target recognition. In this paper, a novel approach is proposed to accurately estimate the micro-Doppler frequencies of specific target scattering parts, which can be very promising in real-time human sensing applications. For this approach, the adapted linear predictive coding algorithm is first applied to extend the known data in each time window. Then, the Hough transform with a set of new fitting models is used to extract the interested echo components and accurately estimate their micro-Doppler frequencies. Finally, a series of experiments are conducted to illustrate the performance of the proposed approach.

Are idiosyncrasies in vowel production free or learned? A study of variants of the French vowel system in biological brothers

Speech production displays a number of idiosyncrasies that are individual variations in the way speakers achieve phonetic contrasts in their language. It was shown previously [Ménard L. et al., Speech Commun. 2008, 50(1), 14-28] that idiosyncrasies in the production of the height contrast in oral vowels in French are characterized by large variations in the distribution of F1 values, which are associated with a stability of F1 for a given height degree, independent of the place of articulation (front vs. back) and rounding. The current study aimed to assess whether these idiosyncrasies are random or induced by the learning environment. Ten pairs of French Canadian adult male siblings were recruited, and each individual was asked to produce ten repetitions of the ten French oral vowels, which were recorded. F1 values were extracted using linear predictive coding algorithms. There was a trend towards imposed variations, since the distances between F1 values for brothers for a given vowel were significantly smaller than the corresponding distances between speakers who were not brothers. However, the F1 values within pairs of brothers were significantly correlated for only two of the six mid-high or mid-low vowels. Thus, it appears that a large part of the idiosyncrasies in the pronunciation of vowels were random and differed between brothers.

Lossless predictive coding of electric signal waveforms

This paper describes a new predictive coding algorithm designed for lossless compression of electric signal waveforms. Prediction coefficients are obtained from the sinusoidal signal autoregressive model. Predictors for a single sinusoid and the sum of two and three sinusoids are considered. Integer-to-integer coding is obtained by the proper (i.e. reversible) quantization of computations. The proposed coding algorithm shows better performance than the commonly used 2nd order differences coding and linear prediction coding that is optimal in the least squares sense.

A chaotic time series prediction model for speech signal encoding based on genetic programming

In this paper, a novel solving method for speech signal chaotic time series prediction model was proposed. A phase space was reconstructed based on speech signal's chaotic characteristics and the genetic programming (GP) algorithm was introduced for solving the speech chaotic time series prediction models on the phase space with the embedding dimension m and time delay τ. And then, the speech signal's chaotic time series models were built. By standardized processing of these models and optimizing parameters, a speech signal's coding model of chaotic time series with certain generalization ability was obtained. At last, the experimental results showed that the proposed method can get the speech signal chaotic time series prediction models much more effectively, and had a better coding accuracy than linear predictive coding (LPC) algorithms and neural network model.

Acoustic assessment of erygmophonic speech of Moroccan laryngectomized patients.

IntroductionAcoustic evaluation of alaryngeal voices is among the most prominent issues in speech analysis field. In fact, many methods have been developed to date to substitute the classic perceptual evaluation. The Aim of this study is to present our experience in erygmophonic speech objective assessment and to discuss the most widely used methods of acoustic speech appraisal. through a prospective case-control study we have measured acoustic parameters of speech quality during one year of erygmophonic rehabilitation therapy of Moroccan laryngectomized patients.MethodsWe have assessed acoustic parameters of erygmophonic speech samples of eleven laryngectomized patients through the speech rehabilitation therapy. Acoustic parameters were obtained by perturbation analysis method and linear predictive coding algorithms also through the broadband spectrogram.ResultsUsing perturbation analysis methods, we have found erygmophonic voice to be significantly poorer than normal speech and it exhibits higher formant frequency values. However, erygmophonic voice shows also higher and extremely variable Error values that were greater than the acceptable level. And thus, live a doubt on the reliability of those analytic methods results.ConclusionAcoustic parameters for objective evaluation of alaryngeal voices should allow a reliable representation of the perceptual evaluation of the quality of speech. This requirement has not been fulfilled by the common methods used so far. Therefore, acoustical assessment of erygmophonic speech needs more investigations.

유도전동기의 고장 진단을 위한 효과적인 특징 추출 방법

본 논문은 고장 분류 시스템을 위해 진동 신호로부터 특징 벡터를 자동적으로 추출하는 효과적인 기법을 제안한다. 기존의 멜-주파수 캡스트럼 계수는 진동신호의 노이즈에 민감하여 분류 정확도를 감소시키는 단점이 있다. 이러한 문제를 해결하기 위해 본 논문은 4단계 필터 뱅크로 구성된 스펙트럴 엔벨로프 캡스트럼 계수 분석을 제안하며, 4단계는 (1) 모든 진동 신호의 스펙트럴 엔벨로프를 기술하기 위한 선형 예측 코딩 알고리즘 사용 단계, (2) 일반적인 스펙트럴 모양을 얻기 위해 모든 엔벨로프의 평균화 단계, (3) 평균 엔벨로프와 그 주파수의 최대값을 찾기 위한 기울기 하강 방법 사용 단계, (4) 엔벨로프의 주파수 사이의 거리로부터 계산된 중앙값을 얻는데 사용되는 비 중첩 필터 뱅크 단계로 구성된다. 이4-단계필터뱅크는 특징벡터를 추출하기위해 캡스트럼 계수 계산에 사용된다. 마지막으로 유도전동기의 결함 형태를 구분하기 위해 이러한 특수 파라미터를 사용하는 다중 계층 서포트 벡터 머신을 사용한다. 모의실험 결과, 제안하는 방법은 약 99.65%의 분류 성능을 보이며, 동시에 기존 방법들보다 우수한 성능을 보인다. This paper proposes an effective technique that is used to automatically extract feature vectors from vibration signals for fault classification systems. Conventional mel-frequency cepstral coefficients (MFCCs) are sensitive to noise of vibration signals, degrading classification accuracy. To solve this problem, this paper proposes spectral envelope cepstral coefficients (SECC) analysis, where a 4-step filter bank based on spectral envelopes of vibration signals is used: (1) a linear predictive coding (LPC) algorithm is used to specify spectral envelopes of all faulty vibration signals, (2) all envelopes are averaged to get general spectral shape, (3) a gradient descent method is used to find extremes of the average envelope and its frequencies, (4) a non-overlapped filter is used to have centers calculated from distances between valley frequencies of the envelope. This 4-step filter bank is then used in cepstral coefficients computation to extract feature vectors. Finally, a multi-layer support vector machine (MLSVM) with various sigma values uses these special parameters to identify faulty types of induction motors. Experimental results indicate that the proposed extraction method outperforms other feature extraction algorithms, yielding more than about 99.65% of classification accuracy.

The Research for Compression Algorithm of Aerial Imagery

In this study, the new method of the JPEG image compression algorithm with predictive coding algorithm combining was proposed, effectively eliminates redundant information of the sub-blocks and redundant information between the sub-blocks and sub-blocks. Achieved higher compression ratio compared to the JPEG compression algorithm and a good image quality.

Contribution of Spectral Cues to Mandarin Lexical Tone Recognition in Normal-Hearing and Hearing-Impaired Mandarin Chinese Speakers

The purpose of this study was to investigate the contribution of spectral fine structure and spectral envelope cues to recognition of Mandarin lexical tones in normal-hearing and sensorineural hearing-impaired Mandarin-speaking listeners. Four groups of subjects participated in the study, including 20 normal-hearing, 20 moderately, 20 moderately to severely, and 8 severely hearing-impaired listeners. The original speech materials consisted of 16 sets of Mandarin monosyllables spoken by a male and a female. Each monosyllable had four tonal patterns, resulting in a total of 64 combinations of consonants, vowels, and tones. A Linear Predictive Coding (LPC) algorithm was used to create two sets of synthesized materials, including 128 tokens with the original spectral fine structure mixed with the spectral envelope from a different tone, as well as 128 tokens with noise fine structure and the original spectral envelope. All subjects participated in tone recognition tests using the two sets of chimeric tone tokens. Oral responses to tones were recorded and scored as percent correct. Hearing-impaired listeners could take advantage of spectral fine structure in the recognition of lexical tones, but with increasing hearing loss, the ability of hearing-impaired listeners to recognize tones became worse, especially for severely hearing-impaired listeners. Hearing-impaired listeners showed significant differences in tone recognition between the male and female voices. Tone 3 was the easiest tone to perceive, followed by tone 2, whereas tones 1 and 4 were hard for all subjects, particularly when only the spectral envelope cue was available. Hearing-impaired listeners showed a significantly lower level of lexical tone recognition than normal-hearing listeners when using spectral envelope cues in comparison with normal-hearing listeners. These results demonstrate that the spectral fine structure cue dominates lexical tone recognition for all subjects. Listeners with sensorineural hearing impairment showed reduced ability in the recognition of lexical tones using both spectral fine structure and spectral envelope cues, which may result from their impaired auditory spectral resolution.

A wavelet implementation of the pioneering block-based disparity compensated predictive coding algorithm for stereo image pair compression

In this paper, we propose a novel, discrete wavelet transform (DWT) domain implementation of our previously proposed, pioneering block-based disparity compensated predictive coding algorithm for stereo image compression. Under the present research context we perform predictive coding in the form of pioneering block search in the sub-band domain. The resulting transform domain predictive error image is subsequently converted to a so-called wavelet-block representation, before being quantized and entropy coded by a JPEG-like CODEC. We show that the proposed novel implementation is able to effectively transfer the inherent advantages of DWT-based image coding technology to efficient stereo image pair compression. At equivalent bit rates, the proposed algorithm achieves peak signal to noise ratio gains of up to 5.5 dB, for reconstructed predicted images, as compared to traditional and state of the art DCT and DWT-based predictive coding algorithms.

Lossy predictive coding of SAR raw data

In this paper, we propose to employ predictive coding for lossy compression of synthetic aperture radar (SAR) raw data. We exploit the known result that a blockwise normalized SAR raw signal is a Gaussian stationary process in order to design an optimal decorrelator for this signal. We show that, due to the statistical properties of the SAR signal, an along-range linear predictor with few taps is able to effectively capture most of the raw signal correlation. The proposed predictive coding algorithm, which performs quantization of the prediction error, optionally followed by entropy coding, exhibits a number of advantages, and notably an interesting performance/complexity trade-off, with respect to other techniques such as flexible block adaptive quantization (FBAQ) or methods based on transform-coding; fractional output bit-rates can also be achieved in the entropy-constrained mode. Simulation results on real-world SIR-C/X-SAR as well as simulated raw and image data show that the proposed algorithm outperforms FBAQ as to SNR, at a computational cost compatible with modern SAR systems.

A new video compression algorithm for different videoconferencing standards

AbstractIn this paper, a new and simple predictive coding algorithm for video compression and multimedia communication for different network applications is investigated. This algorithm is called semi‐hexagonal absolute moment block truncation coding (SH‐AMBTC). It enjoys better objective and subjective qualities over the standard square shaped AMBTC. This technique based on the prediction of the bit‐map of the middle frames in a sequence from the bit‐map of the end frames (the first and the last frames in the group). It is of very low complexity comparing to the standard algorithm MPEG. The new algorithm adds no complexity to the existing AMBTC algorithm. Copyright © 2002 John Wiley & Sons, Ltd.