BaNa: A Noise Resilient Fundamental Frequency Detection Algorithm for Speech and Music

Abstract

Abstract-Fundamental frequency (F0) is one of the essential features in many acoustic related applications. Although numerous F0 detection algorithms have been developed, the detection accuracy in noisy environments still needs improvement. We present a hybrid noise resilient F0 detection algorithm named BaNa that combines the approaches of harmonic ratios and Cepstrum analysis. A Viterbi algorithm with a cost function is used to identify the F0 value among several F0 candidates. Speech and music databases with eight different types of additive noise are used to evaluate the performance of the BaNa algorithm and several classic and state-of-the-art F0 detection algorithms. Results show that for almost all types of noise and signal-to-noise ratio (SNR) values investigated, BaNa achieves the lowest Gross Pitch Error (GPE) rate among all the algorithms. Moreover, for the 0 dB SNR scenarios, the BaNa algorithm is shown to achieve 20% to 35% GPE rate for speech and 12% to 39% GPE rate for music. We also describe implementation issues that must be addressed to run the BaNa algorithm as a real-time application on a smartphone platform.

Utilization of windowing effect and accumulated autocorrelation function and power spectrum for pitch detection in noisy environments

“…In this case, even if the speech signal is corrupted by noise, the affect of noise is suppressed. This is the motivation in [14,15], where a long length of N is used commonly. In this paper, we consider to use the Rectangular window instead of the Hanning and Hamming windows, and keep a standard length of N .…”

“…Reliable detection of the pitch period (T 0 ) being the inverse of the fundamental frequency (F 0 ) from speech is required in a wide range of applications such as speech coding, speech recognition, speech enhancement, speech synthesis, and so on. Therefore, a large number of pitch detection methods have been addressed up to now [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19].…”

“…This is because the Hanning and Hamming windows are adequate for the purpose of keeping a good balance between sharp peaks of speech harmonics and noise suppression except for speech harmonics. Thus, even in the recent techniques of PEFAC [14] and BaNa [15], these windows are used. However, in both techniques, a comparatively longer frame length is proactively utilized.…”

Estimation of Fundamental Frequency from Singing Voice Using Harmonics of Impulse-like Excitation Source

“…It is known that source filter decomposition through LP analysis fails drastically, for the voices such as singing voice, due to significant coupling between source and system. Some methods combine various techniques such as combination of harmonic ratios and cepstral analyses [18,19]. In methods that use time-frequency domain, the speech signal is decomposed into multiple frequency bands and then time domain methods such as autocorrelation are applied on each of the sub bands [20].…”

Robust Estimation of Fundamental Frequency Using Single Frequency Filtering Approach

“…Examples of this kind are sub-harmonics to harmonics ratio (SHRP) [11], summation of residual harmonics [12], dominant harmonics [13], sawtooth waveform inspired pitch estimator (SWIPE) [14], etc. Also, some approaches [15,16] combine various techniques of F0 estimation (e.g., [15] combines harmonic ratios and cepstral analysis) for robustness under degraded conditions.…”