REVIEWS OF ACOUSTICAL PATENTS

• GEORGE L. AUGSPURGER, Perception, Incorporated, Box 39536, Los Angeles, California 90039

• ERIC E. UNGAR, Acentech, Incorporated, 33 Moulton Street, Cambridge, Massachusetts 02138

Ryan Termeulen, assignor to Bose Corporation

12 December 2017; filed 9 November 2016

In this Bose patent, each earphone (or earbud) of a pair houses an array of two or more microphones to pick up exterior sounds. Signals from both sets of microphones are processed to generate a single far-field pickup pattern, “…that is more sensitive to sounds originating a short distance away from the apparatus than to sounds close to the apparatus above the cutoff frequency, and omnidirectional below the cutoff frequency.” Below the cutoff frequency, signals from one array are inverted before left and right signals are combined, thus attenuating low frequencies. The level of wind noise in the microphone signals is calculated and used to adjust the cutoff frequency. A second processor generates a near-field pattern to pick up the user's voice. Various calculations, procedures, and refinements are clearly explained in the patent.—GLA

Hirotaka Sawa et al., assignors to PANASONIC INTELLECTUAL PROPERTY MANAGEMENT CO., LTD.

2 January 2018; filed 14 February 2014

This very long patent consists mostly of huffing and puffing. It finally fizzles out in one very short Claim. Consider a typical surveillance camera that can pan and tilt to aim at a series of preset coverage areas. The camera is linked to a steerable microphone array that can be aimed at the same areas. The patent Claim sets forth a possible method for checking and aligning the aiming points of the two devices.—GLA

Teppei Fukuda et al., assignors to PANASONIC INTELLECTUAL PROPERTY MANAGEMENT CO., LTD.

2 January 2018; filed 9 December 2015

This patent is a companion of U.S. Patent No. 9860439 reviewed above. Both patents are concerned with the installation and calibration of audio-video surveillance systems. In this case, a surveillance camera and a steerable microphone array are mounted some distance apart. Obviously, geometrical calculations could be made to automatically aim the camera at a location detected by the microphone array. However, the patent recommends a manual calibration procedure involving video displays of predetermined locations.—GLA

Michael L. Strange et al., assignors to CORE BRANDS, LLC

2 January 2018; filed 1 May 2015

Audio engineers have long debated whether a loudspeaker should be resiliently attached to its mounting panel, thus minimizing the chance that the panel will act as a sounding board for mechanically transmitted vibrations from the speaker frame. This patent argues that a wall-mounted or ceiling-mounted speaker will benefit from resilient mounting. The patent describes a fairly elaborate arrangement that includes a “clamp damper” and a “flange damper.” —GLA

Alexander Yakovlevich Bogdanov, assignor to Alexander Yakovlevich Bogdanov

2 January 2018; filed 27 September 2013

Forty years ago this patent would have been interesting. It explains (correctly) that when electronic bass boost is used to extend the low frequency response of a small loudspeaker system, the boost should not extend below the usable operating range. For example, a small woofer in a 15-liter box might be equalized down to 45 Hz, with a rolloff below 40 Hz to prevent excessive cone excursion. Today, almost any combination of boost and cutoff is easy to implement with small, inexpensive active filters. However, in the early days of transistorized amplifiers there was some interest in tinkering with feedback in the power amplifier itself, and that is the approach taken here. A bandpass filter is used in a positive feedback loop to achieve the desired electronic response. The patent includes helpful suggestions for maintaining stability under various load conditions, but the interaction between load impedance and frequency response is shrugged off as a minor complication.—GLA

Yasuhiro Miki, assignor to KYOCERA CORPORATION

19 December 2017; filed 22 November 2016

Most of the early portable listening devices used a miniature phone jack for headphone connections, and it became a default standard. The old telephone contact designations of “tip,” “ring,” and “sleeve” became “left,” “right,” and “common.” When smartphones were developed, a fourth contact was added for a microphone connection, and it became possible for at least two different kinds of plugs to be inserted into a headphone jack. Several patented methods were developed to identify what kind of device was plugged in and automatically supply the desired functions. This Kyocera patent points out that some headset transducers require external power, and so a fifth contact must be squeezed in. A combination of physical and electronic features is said to achieve compatibility with all three types of phone plugs.—GLA

Ming-Hung Chang, assignor to ANPEC ELECTRONICS CORPORATION

26 December 2017; filed 20 April 2017

This patent discloses another in a growing list of methods for achieving compatibility between the different kinds of plugs that may be inserted into the headphone jack of a portable listening device. In this case, operational amplifiers for left and right channels are connected to a detection and control circuit which “…controls the first operational amplifier and the second operational amplifier to be in a Hi-Z state and determines the type of the headset according to the voltage at the fourth pin.” It is doubtful whether the pop-free feature can be maintained in the case of a partial or intermittent plug connection.—GLA

Joseph Sylvester Chang et al., assignors to Advanced Electroacoustics Private Limited

2 January 2018; filed 20 March 2012

The preferred embodiment described in this patent's abstract and shown in the illustration is simple: the microphone input of a portable music device can be electrically connected to an earphone output and used for a variety of purposes, such as monitoring noise exposure. However, what is actually patented in the Claims is a different and more complicated arrangement in which the electrical signal that normally drives one earphone is instead used to energize an induction coil that can then be picked up by a hearing aid, “…without requiring a battery or external power source.”—GLA

John Timothy and Sam Noertker, assignors to Skullcandy, Inc.

2 January 2018; filed 29 December 2015

This patent describes a gaming headset in which each earcup houses a conventional loudspeaker plus a separate tactile vibrator. The vibrator is intended to enhance the wearer's perception of very low frequencies. In this design, both transducers are fed from a single power amplifier but their relative levels are balanced by adding resistors 138 between the amplifier and the loudspeaker. That's right—the inventors have managed to patent a series resistor. Not very efficient and not very original, but worth a U.S. patent.—GLA

Takeshi Hashimoto et al., assignors to CLARION CO., LTD.

9 January 2018; filed 20 February 2015

Audio systems installed in a car seat often include a subwoofer imbedded in the seat or backrest. Such a subwoofer must generate significant levels of sound, since this sound is transmitted to the occupant via the seat structures, which provide considerable attenuation. This patent addresses means for obtaining a usefully high signal without excessive power consumption. It describes data processing means that enable the transmission of lower power signals by taking advantage of the amplifications resulting from resonances in the system.—EEU

Werner Paulus Josephus De Bruijn et al., assignors to KONINKLIJKE PHILIPS N.V.

2 January 2018; filed 6 May 2014

A great deal of effort has been aimed at finding efficient ways to reproduce surround sound in home installations with non-optimal loudspeaker placement, and several methods have been standardized. The new method proposed in this patent can select or combine various rendering modes. It seems best suited to installations with a large number of loudspeakers, and there must be a separate power amplifier for each loudspeaker. The novel feature is that loudspeakers are grouped into clusters for signal processing. It is possible for a cluster to contain only one loudspeaker. It is also possible for a loudspeaker to be assigned to more than one cluster. “The apparatus is configured to select array processing techniques for specific subsets that contain audio transducers that are sufficiently close and allow automatic adaptation to audio transducer configurations, thereby allowing a user increased flexibility in positioning loudspeakers.”—GLA

Dirk Jeroen Breebaart et al., assignors to KONINKLIJKE PHILIPS N.V.

9 January 2018; filed 6 April 2015

This patent describes (at considerable length) yet another method for compressed encoding of surround sound. The invention is said to be especially well suited for converting a multi-channel signal to a binaural spatial signal.—GLA

Adesina Akeem Yusuf et al., assignors to King Fahd University of Petroleum and Minerals

26 December 2017; filed 1 September 2015

The natural frequency of a plate with a welded seam, and that of a similar uniform plate (without a seam), are determined by subjecting each of the plates to impacts. The uniformity of the weld is deduced by comparing the natural frequencies of the two plates.—EEU

Joseph W. Pillis et al., assignors to Johnson Controls Technology Company

2 January 2018; filed 7 August 2015

Vibrations of a screw compressor are reduced by means of a squeeze film damper that is arranged in parallel with the compressor's radial or thrust bearings. This arrangement provides damping while maintaining the alignments of the rotating parts. The damping fluid conveniently may be the same as the lubricant used in the radial or thrust bearings.—EEU

Antonio Sangermano, II et al., assignors to Bose Corporation

26 December 2017; filed 31 March 2016

This patent describes, in essence, a demonstration seat that can be placed in a vehicle to give potential users an idea of how various actively isolated seats would perform. The demonstration seat includes actuators that respond to controllers which can be programmed so that the system mimics seats with various active vibration isolation configurations.—EEU

Ravi K. Sharma et al., assignors to Digimarc Corporation

26 December 2017; filed 4 April 2016

Techniques are presented for recognizing various types of miscellaneous information in an acoustic signal, including background noises, such as traffic or music, as well as adding signal content to an audio signal, possibly in a concealed form, and then extracting such content from the processed signal. The process known as watermarking deals with concealing information within an audio signal with minimal effect of the final audio stream.—DLR

Michiel van Nieuwkerk and Yang Gao, assignors to Sonion Nederland B.V.

9 January 2018; filed 25 January 2017

Tiny integrated circuit amplifiers used in hearing aids must draw as little current as possible. According to this patent, a two-wire self-biasing arrangement is commonly used, and a high-pass filter is usually incorporated into the amplifier design. We are told that the high-pass response of prior art designs can be affected by signal level. An improved circuit is disclosed which is said to avoid this problem and offer other advantages as well.—GLA

Tadashi Yamaura, assignor to BlackBerry Limited

26 December 2017; filed 12 February 2016

Certain improvements in speech coding and transmission are discussed in the patent. The coding technique known as code-excited linear prediction, or CELP, was described several decades ago. This patent discusses methods for reducing the need for sending new excitation codebooks, using new ways of analyzing the noise content of the signal.—DLR

Phillip L. De Leon et al., assignors to VoiceCipher, Inc.

9 January 2018; filed 21 August 2014

The patent discusses issues involved in the detection of synthetic speech when it is used in an attempt to mimic a human speaker for access to secure systems using voice identification methods. Current speech synthesis technology does not properly mimic a number of aspects of a human speech signal. These discrepancies can be detected by a spoofing detector system, such as described. In particular, differences are detected in pitch frequency stability within phonemes and across phoneme boundaries. The figure shows differences between human and synthetic speech for pitch jitter measurements of several common words, as measured across many speakers.—DLR

Anurag Ratan Singh et al., assignors to Nuance Communications, Inc.

26 December 2017; filed 8 June 2015

The patent deals with the pronunciation of cross-language proper nouns, particularly people's names and place names, by a speech synthesis system. Such issues as syllable stress and vowel reduction are some of the more common problems that arise in the pronunciation of foreign words. Names are particularly difficult in these respects. Various sources may be consulted to help make such decisions.—DLR

Gaurav Talwar et al., assignors to GM Global Technology Operations LLC

9 January 2018; filed 22 March 2016

The described system would attempt to improve the quality of synthesized speech by performing one or more speech recognition-type steps on the synthesized speech before playing that speech to the user. The primary recognition-like operation mentioned is a hidden Markov model (HMM) analysis. The HMM results would be processed much as if incoming speech were being analyzed. Details of the HMM analysis can then be used to evaluate the quality of the synthesized speech.—DLR

Peter Alex Korn, assignor to Amazon Technologies, Inc.

9 January 2018; filed 29 September 2014

This patent deals with the issue of how secondary information, such as footnotes, should be handled during the audio presentation of on-line documents, such as when a webpage is read out by a text-to-speech system. For example, the system might play a short beep when the footnote was first encountered. The user could then use a gesture, or other input, to signal that the system should now go back and present the footnote or other side information that was earlier skipped over.—DLR

Hsun-Fu Liu et al., assignors to ASUSTeK COMPUTER INC.

9 January 2018; filed 2 December 2015

The patent describes a multi-lingual text-to-speech (TTS) system. Specifically, a TTS system for reading out a web page in spoken form may encounter materials on the page in a second language. The patent discusses the extent to which it might be possible, or even desirable, to use a single phoneme pronunciation database for reproducing sounds from two or more languages. Several issues of cross-language blending are discussed.—DLR

Hyun-Jun Kim and Young Sang Choi, assignors to Samsung Electronics Co., Ltd.

12 December 2017; filed 6 February 2015

The patented speech recognition system goes through the usual steps of acoustic, phonetic and semantic analysis, but at all steps, it uses the recognition results from two separate analysis systems, one which includes current context information and one which does not include such information. The similarities and differences in the two analyses are said to provide better recognition results.—DLR

Michael George, assignor to SANOFI-AVENTIS DEUTSCHLAND GMBH

12 December 2017; filed 18 May 2011

The premise put forward in the description of this patented medical procedures device is that an automated speech interaction with the patient may provide sufficient information to proceed with medical treatment, including giving injections or other such treatments that would normally require a trained medical professional. The patent includes a number of provisions for dealing with recognition errors and patient non-familiarity with materials or procedures. The device could include patient-specific knowledge or patient-group abilities, such as language usage or regional customs. Human medical assistance would be easily available during the procedures.—DLR

Fadi Biadsy and Pedro J. Moreno Mengibar, assignors to Google Inc.

12 December 2017; filed 12 February 2014

The patent describes a speech recognition method based on the idea that non-linguistic information may be of value in making decisions about a speaker's word choices and in the correct interpretation of the words actually spoken. The patent includes examples of what is intended by describing a recognition language model as using a non-linguistic context. For example, a cell phone may be able to detect the user's location or state of motion and use that information to make better recognition decisions.—DLR

Kang-eun Lee et al., assignors to SAMSUNG ELECTRONICS CO., LTD.

19 December 2017; filed 14 June 2016

The patent deals with the effects of the transmission of speech acoustic signals through body parts, such as wrist and hand structures, on the extraction of the signal characteristics, such as voicing and formant frequencies, as required for acceptable speech recognition performance.—DLR

Brian T. Lillard and Karl H. Koster, assignors to NOBLE SYSTEMS CORPORATION

19 December 2017; filed 28 March 2016

The patent describes a voice processing system that would be used by a call center to take initial caller information and provide initial feedback to the caller, and would also begin building a contact information form that would be used by a human operator to respond to the caller's inquiry. Initial caller feedback could be in the form of speech or a text to a user's device.—DLR

Adam Jordan et al., assignors to PROMPTU SYSTEMS CORPORATION

19 December 2017; filed 16 December 2014

The invention is intended to address a number of issues that arise during voice control of media devices, such as television program navigation and viewing. Many of the issues the viewer must deal with are addressed in a general manner, hopefully in such a way that the invention would be applicable to a wide variety of such systems. Several interaction scenarios are described, such as calling up a specific movie for viewing.—DLR

Katharine Nielson et al., assignors to VOXY, Inc.

26 December 2017; filed 12 February 2016

The patent describes a number of the problems faced by a language learner and how such problems may be addressed in user-specific and context-specific ways. The patented system would use both language-related knowledge and user-specific information, such as the current state of the user's language ability, to tailor the presentation of materials in the language being taught.—DLR

Eugene Weinstein et al., assignors to Google Inc.

2 January 2018; filed 23 June 2014

With increasing use of small, portable user devices for speech recognition tasks, there is new interest in moving the computational load to remote devices, such as servers and other off-site computers. This patent discusses ways of using alternate resources, including access to prior knowledge, as a substitute for fully processing the incoming speech signal.—DLR

Denis Sergeyevich Filimonov et al., assignors to Amazon Technologies, Inc.

9 January 2018; filed 20 June 2016

The patent discusses the use of several types of finite-state transducers (FSTs) for the recognition of speech in applications where immediate extraction of semantic information is required, such as control of household objects by speech. An FST consists of a sequence of elements, phonemes, words, etc., and allows the likelihood of subsequent elements to be predicted. Occurrence of a particular sequence may thus be taken as an indicator of the recognized utterance. These models are used in currently available recognition systems, such as devices marketed by the inventor.—DLR

Kazuya Nomura, assignor to PANASONIC INTELLECTUAL PROPERTY CORPORATION OF AMERICA

9 January 2018; filed 8 July 2014

Speech recognition devices are increasingly being used to control household devices, such as lighting and media operation, channel switching, sound control, etc. Typically, such devices have required either a button press or the utterance of a particular keyword in order to start the recognizer system. This patent suggests that such control methods can be made more natural if the device actually recognizes everything it hears and makes its own decisions about whether or not it is to act on the recognized speech content. Some of the pros and cons of this approach are discussed.—DLR

Sean A. Ramprashad et al., assignors to APPLE INC.

9 January 2018; filed 6 June 2015

Multiple microphones have often been used as a way to capture more audio information which may be useful in a speech recognition system. Two quite different microphone arrangements have equally different effects on the system design. In one case, such as in the Amazon Alexa system, the microphones are all located within a small space. This allows the use of techniques known as beamforming, allowing the speech source direction to be determined. This patent discusses a different technique in which the microphones are placed at different locations throughout the area being monitored. This allows the comparison of different audio streams and the detection of speech originating at a particular location in the space.—DLR

Michael R. Sumner et al., assignors to APPLE INC.

9 January 2018; filed 28 August 2015

The patented speech recognition system would use so-called “metadata” to arrive at the final recognition output. This metadata would consist of other actions being taken by the user, such as placing a cursor at a particular blank on a form, or perhaps awareness of the environment, the user's current location or motion, or other activities. Other such information could include the recent history or longer-term knowledge of the user's style and activity patterns.—DLR

Yaakov Chen and Doron Koren, assignors to DSP GROUP LTD.

9 January 2018; filed 2 June 2016

According to the patent, whispered speech has no periodic excitation or harmonic structure, has formant frequencies shifted higher than normal, has a flatter spectral slope and typically longer durations. It is stated clearly that the patented method would not detect whispered speech by audio processing alone, but that various additional clues are required. These include the mouth being very close to the microphone and possibly shielded by a hand. Analysis techniques may include the usual list of methods, such as hidden Markov models, neural networks, dynamic time warping, mel cepstra, etc. Basics such as gain control and noise detection may be used in unique ways.—DLR