Prior art

To identify the most relevant prior art for US patent 9,031,259, I will first access the patent document itself to view its cited references. The USPTO provides tools for patent searching, including Patent Center and Patent Public Search.

As I do not have direct access to the USPTO database for live searching, I will refer to the full patent text provided in the prompt to extract the cited prior art. The full patent text mentions two specific prior art references within the description of the speech segment determiner 15: U.S. patent application Ser. No. 13/302,040 and U.S. patent application Ser. No. 13/364,016.

Here's an analysis of the potentially anticipatory prior art mentioned in the patent text:

Prior Art Reference 1:

• Full Citation: U.S. patent application Ser. No. 13/302,040
• Publication/Filing Date: The patent text does not explicitly state the publication or filing date for this application number, but it is referenced within the context of US9031259's filing date of September 14, 2012, indicating it precedes the present invention.
• Brief Description: This patent application describes "speech segment determination technique I," which focuses on the frequency spectra of vowel sounds to detect speech segments. It involves obtaining a signal-to-noise ratio (SNR) between the peak level of a vowel-sound frequency component and a noise level in each frequency band. A speech segment is detected if the obtained SNR meets a specific ratio for a specific number of peaks. The speech segment determiner 15a (FIG. 2) in US9031259 employs this technique, utilizing components like a frame extraction unit, spectrum generation unit, subband division unit, frequency averaging unit, storage unit, time-domain averaging unit, peak detection unit, and speech determination unit.
• Potential Anticipation (35 U.S.C. § 102): This reference potentially anticipates aspects of the speech segment determiner described in the independent claims. Specifically, it could anticipate the step of "determining whether or not a sound picked up by at least either a first microphone or a second microphone is a speech segment" (as recited in the Noise Reduction Apparatus and Audio Input Apparatus claims) and the "determining whether or not a sound picked up by at least either a first microphone or a second microphone is a speech segment" step in the Noise Reduction Method claim. The detailed method of detecting speech segments based on vowel sound characteristics, subband energy, and SNR calculation, as described in US9031259 with reference to this application, is likely taught by this prior art.

Prior Art Reference 2:

• Full Citation: U.S. patent application Ser. No. 13/364,016
• Publication/Filing Date: The patent text does not explicitly state the publication or filing date for this application number, but it is referenced alongside the first prior art, indicating it also precedes the present invention.
• Brief Description: This patent application describes "speech segment determination technique II," which focuses on the characteristics of consonants, specifically their spectral patterns that tend to rise to the right. It detects a consonant segment in an intermediate to high-frequency band by extracting a frequency distribution of the consonant that is less affected by noise. The speech segment determiner 15b (FIG. 3) in US9031259 uses this technique, involving a frame extraction unit, spectrum generation unit, subband division unit, average-energy derivation unit, noise-level derivation unit, determination-scheme selection unit, and consonant determination unit. It includes different determination schemes based on noise levels.
• Potential Anticipation (35 U.S.C. § 102): Similar to the first reference, this prior art potentially anticipates aspects of the speech segment determiner as claimed in US9031259. It would specifically relate to the "determining whether or not a sound picked up by at least either a first microphone or a second microphone is a speech segment" clause within the independent claims, by teaching a specific and robust method for consonant-based speech detection. The detailed comparison of subband average energy between consecutive subbands to identify consonant characteristics is likely taught by this prior art.