Acoustic Processing Unit Interface for Real-Time Speech Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Speech recognition systems place a significant load on computing resources, affecting processing speed and memory usage, especially in real-time applications and handheld devices with limited resources.
Innovation Solution
A dedicated Acoustic Processing Unit (APU) is introduced, which divides audio signals into frames, compares feature vectors to Gaussian probability distributions, and outputs scores, while the Central Processing Unit (CPU) performs concurrent search operations, reducing the computational burden on the CPU and optimizing memory usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If speech recognition processing is performed on the CPU, then speech recognition accuracy is improved, but processing speed and system performance deteriorate due to high computational load
Solution Approach 1:
The patent divides the speech recognition system into two separate processing units: the CPU handles high-level control and coordination, while the APU handles specific acoustic modeling computations. This segmentation allows each unit to specialize in its designated tasks, improving overall system efficiency and processing speed without sacrificing recognition accuracy.
Solution Approach 2:
The APU acts as an intermediary component between the CPU and acoustic model data. It receives audio data from the CPU, performs Gaussian probability distribution computations locally using its dedicated memory, and returns results to the CPU. This intermediary structure reduces the computational burden on the CPU while maintaining accurate speech recognition.
2Adaptability or versatility
If speech recognition processing is performed on the CPU, then comprehensive processing capability is maintained, but memory resources are consumed excessively
Solution Approach 1:
The patent extracts the acoustic model data and computation functionality from the CPU's memory and processing resources, placing them in the APU's dedicated local memory. This extraction reduces the memory burden on the CPU and allows the system to maintain comprehensive processing capability while consuming fewer overall memory resources.
Solution Approach 2:
The APU maintains a local copy of the acoustic model data in its dedicated memory, separate from the CPU's memory space. This copying approach allows the APU to perform computations independently without competing for the CPU's memory resources, thereby reducing overall memory consumption while preserving full processing capability.
3Speed
If real-time speech recognition is implemented, then processing speed is improved, but computational load on the system increases
Solution Approach 1:
The patent segments the computational workload by assigning real-time audio processing and Gaussian probability calculations to the APU, while the CPU handles higher-level tasks. This segmentation enables real-time processing speed without concentrating excessive computational load on a single processor.
Solution Approach 2:
The APU performs self-service by maintaining its own local memory for acoustic model data and executing computations independently without requiring continuous CPU intervention. This self-sufficient design reduces the overall computational load on the system while maintaining real-time processing capability.
Data Source
AI summary
Embodiments of the present invention include an apparatus, method, and system for acoustic modeling. In an embodiment, a speech recognition system is provided. The system includes a processing unit configured to divide a received audio signal into consecutive frames having respective frame vectors, an acoustic processing unit (APU), a data bus that couples the processing unit and the APU. The APU includes a local, non-volatile memory that stores a plurality of senones, a memory buffer coupled to the memory, the acoustic processing unit being configured to load at least one Gaussian probability distribution vector stored in the memory into the memory buffer, and a scoring unit configured to simultaneously compare a plurality of dimensions of a Gaussian probability distribution vector loaded into the memory buffer with respective dimensions of a frame vector received from the processing unit and to output a corresponding score to the processing unit. The APU is further configured to divide a clock frequency associated with the received audio signal to a frequency greater than the clock frequency associated with the received audio signal in order to help the score calculation operate faster than the clock frequency of the received audio signal.