Bit-Serial Neural Template Matching for Real-Time Pattern Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current technologies for detecting neural population patterns in brain activity are computationally intensive and require significant resources, making them unsuitable for real-time applications, especially as advances in probe technology enable sampling more neurons, necessitating a more efficient and portable solution.
Innovation Solution
A system and method for template matching that utilizes a processor-based architecture with bit-serial arithmetic units and a hardware-efficient template encoding method, performing operations such as element-wise multiply-sum, accumulation, and sum of squares using a time-dependent sliding indicator window to determine a Pearson's Correlation Coefficient (PCC) value, which is then used to match neural signal streams with pre-recorded templates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional computational methods are used for neural population pattern detection, then detection accuracy can be maintained, but computational resources and processing time increase significantly
Solution Approach 1:
The patent segments the neural population data into discrete time bins and represents neural activity as binary spike indicators (0 or 1) for each neuron in each bin. This segmentation transforms continuous neural signals into discrete, manageable units that can be processed efficiently using bit-serial operations, reducing computational complexity while preserving pattern detection capability
Solution Approach 2:
The patent uses template patterns as simplified copies or representations of expected neural activity sequences. Instead of comparing against complex, full-resolution neural recordings, the system uses compressed template vectors that capture the essential pattern characteristics, enabling efficient correlation computation while maintaining detection accuracy
2Adaptability or versatility
If more neurons are sampled using advanced probe technology, then detection capability improves, but computational load and resource requirements increase
Solution Approach 1:
The patent replaces traditional floating-point arithmetic operations with bit-serial binary operations. By substituting complex mechanical computation (floating-point math) with simpler digital logic operations (bit-wise AND, OR, XOR, and shift operations), the system can process data from increasingly large neural populations without proportionally increasing power consumption
Solution Approach 2:
The patent changes the numerical representation parameters from floating-point to binary integer format. Neural spike trains are encoded as sequences of 0s and 1s, and template patterns are stored as binary vectors. This parameter change enables the use of efficient bit-serial arithmetic units that consume significantly less power than traditional floating-point processors, allowing scalable processing of large neural populations
3Adaptability or versatility
If real-time processing is implemented for portable devices, then application versatility improves, but computational speed and processing efficiency must be optimized
Solution Approach 1:
The patent implements periodic binning of neural spike data into fixed time windows (e.g., 10ms bins). This periodic structure allows the system to process neural data in regular, predictable intervals using identical computational routines, enabling efficient pipelining and real-time operation on portable devices with limited processing power
Solution Approach 2:
The patent performs preliminary encoding of template patterns into binary format and pre-computes correlation metrics before actual neural data arrives. By preparing template vectors and their associated computational parameters in advance, the system eliminates complex computations during real-time processing, achieving faster processing speeds suitable for portable applications
Data Source
AI summary
There is provided a system and method for template matching for neural population pattern detection. The method including: receiving neuron signal streams and serially associating a bit indicator with spikes from each neuron signal stream; serially determining a first summation (S1), a second summation (S2), and a third summation (S3) on the received neuron signals, the first summation including an element-wise multiply-sum using a time-dependent sliding indicator window on the received neuron signal streams and a template, the second summation including an accumulation using the time-dependent sliding indicator window, and the third summation including a sum of squares using the time-dependent sliding indicator window; and determining a correlation value associated with a match of the template with the received neural signal streams, the correlation value determined by combining the first summation, the second summation, and the third summation with predetermined constants associated with the template.