Hidden Markov Model Processing Engine for Diverse Structures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing Hidden Markov Model (HMM) processing engines are limited to handling only one type of HMM structure, such as Ergodic, left-to-right, or parallel path left-to-right models, making them inefficient for processing diverse HMM applications and exacerbating computational intensity with high-state models.
Innovation Solution
A processing engine that receives HMM information, processes back pointer data and state scores for NULL and non-NULL states, updates scores based on predecessor states, and transfers updated scores to an external system, enabling efficient handling of various HMM structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If HMM processing engines are designed to handle only one type of HMM structure, then the processing logic is simplified, but the adaptability to different HMM applications is reduced
Solution Approach 1:
The processing engine is designed to handle multiple types of HMM structures (Ergodic, left-to-right, and parallel path left-to-right models) through a unified architecture. The engine uses a common data structure with state type flags and processing modules that can adapt to different HMM types, allowing one engine to perform multiple functions without requiring separate specialized engines for each HMM structure type.
2Measurement precision
If HMM models with high number of states are processed by traditional engines, then the model accuracy is improved, but the computational intensity increases significantly
Solution Approach 1:
The processing engine divides the computation into distinct segments based on state types (NULL states versus non-NULL states). The engine processes NULL states separately from non-NULL states using different processing logic, and further segments computation by predecessor state relationships. This segmentation allows the engine to efficiently handle high-state models by avoiding redundant computations across all states.
3Productivity
If separate processing engines are created for each HMM structure type, then the processing efficiency for each specific type is optimized, but the overall system complexity and resource requirements increase
Solution Approach 1:
The processing engine employs dynamic processing paths that adapt to the specific HMM structure being processed. The engine uses state type flags and conditional logic to dynamically select appropriate processing methods for different HMM types (Ergodic, left-to-right, parallel path left-to-right). This dynamic approach allows the single engine to achieve processing efficiency comparable to specialized engines while maintaining a unified architecture.
Data Source
AI summary
A method, apparatus, and tangible computer readable medium for processing a Hidden Markov Model (HMM) structure are disclosed herein. For example, the method includes receiving Hidden Markov Model (HMM) information from an external system. The method also includes processing back pointer data and first HMM states scores for one or more NULL states in the HMM information. Second HMM state scores are processed for one or more non-NULL states in the HMM information based on at least one predecessor state. Further, the method includes transferring the second HMM state scores to the external system.