Speech Waveform Alignment via Trigonometric Function Pairing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing speech coding methods face challenges in efficiently aligning periodic speech waveforms, particularly in the frequency domain, due to high computational complexity and the need for precise phase shifts, which can lead to poor results when encoding both periodic and non-periodic speech segments.

Innovation Solution

A method that evaluates trigonometric functions for each phase shift within a range to calculate correlation measures between periodic speech waveforms, using a technique that pairs phase shifts and reduces the number of trigonometric function evaluations, thereby simplifying the alignment process and reducing computational burden.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If cross-correlations are calculated in the frequency domain for prototype alignment, then alignment precision is improved, but computational complexity increases

Engineering Contradiction:
Improvealignment precisionVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the frequency range into multiple segments or bins, calculating cross-correlations for each segment separately. This segmentation allows the use of simpler computational methods within each segment while maintaining overall alignment precision, thereby reducing total computational complexity compared to calculating across the entire frequency range at once.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent calculates cross-correlations only for a subset of frequency components or phase shifts rather than exhaustively computing all possible combinations. By performing partial calculations on selected frequency bins or phase shift values, the method achieves sufficient alignment precision without the full computational burden of complete frequency-domain cross-correlation.

Inventive Principle:
Principle #16Partial or excessive action

2Measurement precision

If multiple phase shifts are evaluated to ensure accurate waveform alignment, then alignment accuracy is improved, but the number of trigonometric function evaluations increases

Engineering Contradiction:
Improvealignment accuracyVSAvoidprocessing efficiency
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent combines the evaluation of multiple phase shifts by computing cross-correlations for paired phase shift values simultaneously or in an integrated manner. By merging the calculation process for multiple phase shifts into a unified computation framework, the method reduces the total number of separate trigonometric function evaluations needed while still achieving accurate alignment through multiple phase shift assessment.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent performs preliminary calculations of trigonometric functions that can be reused across multiple phase shift evaluations. By pre-computing certain trigonometric values or intermediate results that remain constant across different phase shifts, the method reduces redundant calculations and improves processing efficiency while maintaining the ability to evaluate multiple phase shifts for accurate alignment.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS8145477B2Systems, methods, and apparatus for computationally efficient, iterative alignment of speech waveforms
Publication Date: 2012.03.27 QUALCOMM INC
  • US8145477B2 patent drawing
  • US8145477B2 patent drawing
  • US8145477B2 patent drawing

AI summary

Systems, methods, and apparatus described include waveform alignment operations in which a single set of evaluated cosines and sines is used to calculate cross-correlations of two periodic waveforms at two different phase shifts.