Christoph Matthias Nelke

Noise reduction for dual-microphone mobile phones exploiting power level differences

This paper discusses the application of noise reduction algorithms for dual-microphone mobile phones. An analysis of the acoustical environment based on recordings with a dual-microphone mock-up phone mounted on a dummy head is given. Motivated by the recordings, a novel dual-channel noise reduction algorithm is proposed. The key components are a noise PSD estimator and an improved spectral weighting rule which both explicitly exploit the Power Level Differences (PLD) of the desired speech signal between the microphones. Experiments with recorded data show that this low complexity system has a good performance and is beneficial for an integration into future mobile communication devices.

Dual microphone noise PSD estimation for mobile phones in hands-free position exploiting the coherence and speech presence probability

This contribution addresses the enhancement of noisy speech signals picked up by a dual microphone mobile phone in hands-free position. A novel technique to estimate the noise power spectral density is presented which combines two methods: a single microphone algorithm based on the speech presence probability and a dual microphone technique exploiting the coherence properties of the target signal and the background noise. Due to the novel approach, the weakness of both methods can be overcome. Since the proposed method requires knowledge of the current coherence properties, a technique is presented which estimates the coherence of the speech and noise signals, which is usually not known in practice.

Single microphone wind noise PSD estimation using signal centroids

This contribution presents an efficient technique for the enhancement of speech signals disturbed by wind noise. In almost all noise reduction systems an estimate of the current noise power spectral density (PSD) is required. As common methods for background noise estimation fail due to the non-stationary characteristics of wind noise signals, special algorithms are required. The proposed estimation technique consists of three steps: a feature extraction followed by a wind noise detection and the calculation of the current wind noise PSD. For all steps we exploit the different spectral energy distributions of speech and wind noise. In this context, the so-called signal centroids are introduced. Investigations with measured audio data show that our method can cope with the non-stationary characteristics and enables a sufficient reduction of wind noise. In contrast to other wind noise reduction schemes the proposed algorithm has low complexity and low memory consumption.

Measurement, analysis and simulation of wind noise signals for mobile communication devices

In this contribution, we study the characteristics of sound generated by wind and a signal model for the synthesis of wind noise signals is derived. An analysis of the statistics of wind noise recorded in a laboratory setup is carried out with respect to the spectral and temporal properties of the signals. In particular, an autoregresive model is developed for the spectral shape description and the temporal statistics are modeled by a Markov chain. These two components are combined in a model which synthesizes reproducible artificial wind noise signals. Furthermore, a database of measured wind noise signals is provided. The aim of this model and the measured audio data is to provide wind signals for the evaluation of speech enhancement and noise reduction systems.

Wind noise short term power spectrum estimation using pitch adaptive inverse binary masks

This paper presents a method to enhance a speech signal disturbed by wind noise. The wind noise is generated by turbulences in an air stream close to the microphone which picks up the desired speech signal. As the majority of speech enhancement algorithms works in the frequency domain, the short term power spectrum (STPS) of the unwanted noise must be estimated to reduce the wind noise. Conventional algorithms for background noise estimation fail in the case of wind noise due to its non-stationary characteristics. Hence, it is necessary to use special methods for the estimation and reduction of wind noise. The proposed system exploits the spectral characteristics of speech and noise to estimate the wind noise STPS. The spectral power distribution of wind noise and the pitch frequency of speech are used to generate a binary mask for the noise STPS estimation. This method is dependent on a precise pitch estimation. To reduce estimation errors a robust pitch estimation method using knowledge from prior estimates is presented. An evaluation and comparison with other wind noise reduction techniques shows improved speech enhancement of the proposed method.

Corpus based reconstruction of speech degraded by wind noise

This contribution addresses the problem of enhancing a speech signal which is degraded by wind noise. The characteristic that wind noise signals are sparse in time and frequency is exploited in a way that only time-frequency regions that are determined as degraded are enhanced. In these regions of the noisy signal, a process is applied to reconstruct the clean speech data. This is realized by a separation of the noisy speech signal into an autoregressive filter representing the human vocal tract and its excitation signal. The clean filter coefficients of the former are estimated using a pre-trained codebook. A pitch cycle taken from clean speech is adapted to reconstruct the excitation of noisy speech segments.