Hong Kong Applied Science and Technology Research Institute Company, Limited

Kan Ling Capital, L.l.c.

Tao Li

Man Wai Kwan

Yan Huo

Hailiang Li

Pan Zhengang

Yiqing Zhou

Ka Lung Tim Wong

Yu Liu

Wai Ho Mow

Hing-Mo Lam

Xuejiao Liu

Jianhui Wang

Ho Yu Wong

Man Hung Siu

Man Lung Ng

Michael Su

Xiaoqing Shan

A lossless data compressor prevents normalization overruns on-the-fly as symbol occurrence counts are rounded to generate symbol frequencies, allowing an encoding table generator to generate encoding table entries without waiting for the symbol frequency table to finish filling. Rounding errors are accumulated as symbols are normalized and compensated for by reducing a symbol frequency when the symbol frequency is at least 2 and the accumulated errors have exceeded a threshold. The symbol frequency is also reduced when the number of remaining states in the encoding table is insufficient for a number of remaining unprocessed symbols and states for a current encoding table entry. Since error compensation occurs as symbols are being normalized, encoding table generation is not forced to wait for all symbols in the block to be processed, reducing latency. Three pipeline stages can operate on three input blocks: symbol counting, normalization/error compensation/encoding table generation, and data encoding.

Hardware implementation of frequency table generation for Asymmetric-Numeral-System-based data compression

A data-compression analyzer can rapidly make a binary decision to compress or not compress an input data block or can use a slower neural network to predict the block's compression ratio with a regression model. A Concentration Value (CV) that is the sum of the squares of the frequencies and a Number of Zero (NZ) symbols are calculated from an un-sorted symbol frequency table. A rapid decision to compress is signaled when their product CV*NZ exceeds a horizontal threshold THH. During training, CV*NZ is plotted as a function of compression ratio C % for many training data blocks. Different test values of THH are applied to the plot to determine true and false positive rates that are plotted as a Receiver Operating Characteristic (ROC) curve. The point on the ROC curve having the largest Youden index is selected as the optimum THH for use in future binary decisions.

Predicting compression ratio of data with compressible decision

A train-linking lossless data compressor examines a block of data and uses a same coder to generate a same code when all data values in the input block are identical. When the input data is not all the same value, then a Gaussian coder, a Laplace coder, and a delta coder are activated in parallel. The three compressed code lengths are compared and the smallest code length is output as the compressed code when it is smaller than a copy code length. The copy code is a tag followed by copying all the data in the input block. When the smallest of the three compressed code lengths is larger than the copy code length, the file is not compressible, and the copy code is output. No frequency table is required so latency is low. The delta coder subtracts data values from an average value of the last data block.

Train-linking lossless compressor of numeric values

A dynamic-tile neural network accelerator allows for the number and size of computational tiles to be re-configured. Each sub-array of computational cells has edge cells on the left-most column that have an added vector mux that feeds the cell output back to an adder-comparator to allow Rectified Linear Unit (ReLU) and pooling operations that combine outputs shifted in from other cells. The edge cells drive external output registers and receive external weights. The weights and outputs are shifted in opposite directions horizontally between cells while control and input data are shifted in a same direction vertically between cells. A column of row data selectors is inserted between sub-arrays to bypass weights and output data around sub-arrays, while a row of column data selectors are inserted between sub-arrays to bypass control and input data. Larger tiles are configured by passing data directly through these selectors without bypassing.

Dynamic tile parallel neural network accelerator

A synchronizer generates cross-products of In-phase (I) and Quadrature (Q) samples and stores the sign bits for the sine and cosine cross-products. The sign bits are compared to a local reference of a frame-start bit-sequence and the compare results accumulated as I and Q correlations for symbol and half-symbol sampling. Linear combinations of the accumulated I and Q correlations for the symbol and half-symbol sampling generate linear combination results for frequency bins that peak at a different implied Carrier Frequency Offset (CFO) settings. The maximum of the linear combination results is selected and the implied CFO setting for that frequency bin is applied to a demodulator to adjust the receiver's CFO setting and bit synchronization. Computational complexity is reduced since only the sign bit of each cross-product is retained for correlation with the frame-start bit-sequence. Linear combinations can support a wide CFO range.

Cross-product detection method for a narrowband signal under a wide range of carrier frequency offset (CFO) using multiple frequency bins

An Orthogonal Frequency Division Multiplexing (OFDM) receiver comprises: a channel estimator receiving a signal that includes a plurality of symbols, each of the symbols including a plurality of pilots and a plurality of data subcarriers, wherein each symbol further includes at least one pilot subcarrier, the estimator using a first channel response estimate of the at least one pilot subcarrier in a first symbol to select a first weight matrix from a plurality of pre-generated weight matrices, and generating channel estimation information for data subcarriers in the first symbol using the selected first weight matrix.

Channel estimation techniques for OFDM

Systems and methods for time synchronization of orthogonal frequency-division multiplexing (OFDM)-based communications are disclosed. A cyclic prefix (CP) correlation-based timing synchronization method is disclosed for early path (e.g. first-received path) detection for OFDM-based communication. Systems and methods are disclosed for performing double peak value detection in an OFDM-based system for reliably detecting an early (e.g., first-received) path in a received OFDM symbol for use in timing synchronization. In the double peak value detection method, a first peak value detection is performed to detect the path with the largest power, and a second peak value detection is then performed to detect a desired early (e.g., first-received) path. A compensation linear curve is used to facilitate the second peak value detection.

System and method for time synchronization of OFDM-based communications

Joint erasure marking Viterbi algorithm (JEVA), decoder schemes, methods, and systems are provided which perform robust trellis decoder techniques. The provided JEVA decoding schemes are shown to be maximum likelihood decoding schemes that find the most likely transmitted code sequence with a set of symbol erasures, without knowledge of the impulsive noise probability distribution function, and can be implemented sequentially such that they are well suited for dynamically changing impulsive noise channels. In addition, the disclosed details enable various refinements and modifications according to decoder and system design considerations. For example, truncated JEVA is provided for continuous transmission and long frame applications.

Robust joint erasure marking viterbi algorithm decoder

A method comprises receiving a transmission including at least two Orthogonal Frequency Division Multiplexing (OFDM) symbols, estimating a sampling frequency offset associated with the OFDM symbols at least in part by employing a term representing a density of pilots in the OFDM symbols, and compensating for the sampling frequency offset using the estimated sampling frequency offset.

Growing community of inventors

Hong Kong, China

Tao Li